This repository has been archived by the owner on Sep 11, 2020. It is now read-only.
dotgit: ignore filenames that don't match a hash #807
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
For both packfiles and object files. Issue: keybase/client#11366 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu>
strib
force-pushed
the
strib/src-d-ignore-non-hash-files
branch
from
da1231f
to
6ea6016
Compare
|
Can you add a test? |
Suggested by mcuadros. Issue: src-d#807 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu>
|
@mcuadros added a simple test. |
Snehal1112
added a commit
to Snehal1112/go-git
that referenced
this pull request
Dec 1, 2018
* plumbing: format: pktline, Accept oversized pkt-lines up to 65524 bytes The canonical Git client successfully decodes sideband packets up to 65524 bytes in length (4-byte header + 65520-byte payload). The Git protocol documentation was updated in August 2016 to reduce the maximum payload size to 65516 bytes, however old implementations still exist in the wild emitting 65520-byte payloads. As there is no technical difficulty with accepting (not emitting) larger payload sizes, this change adjusts the limit check to allow successful decoding of packets up to 65524 bytes. This change increases compatibility with the current canonical Git implementation. Doc changes from August 2016: git/git@7841c48#diff-52695c8fe91b78b70cea44562ae28297L67 Current packet buffer size is still LARGE_PACKET_MAX (+1 null): https://github.com/git/git/blob/468165c1d8a442994a825f3684528361727cd8c0/sideband.c#L24 https://github.com/git/git/blob/468165c1d8a442994a825f3684528361727cd8c0/sideband.c#L36 LARGE_PACKET_MAX definition: https://github.com/git/git/blob/468165c1d8a442994a825f3684528361727cd8c0/pkt-line.h#L100 Signed-off-by: Joseph Vusich <jvusich@amazon.com> * add PlainOpen variant to find .git in parent dirs This is the git tool's behavior that people are used to; if one runs a git command in a repository's subdirectory, git still works. Fixes src-d#765. Signed-off-by: Daniel Martí <mvdan@mvdan.cc> * use bsd superset for conditional compilation Signed-off-by: wardn <wardn@users.noreply.github.com> * config: adds branches to config for tracking branches against remotes, updates clone to track when cloning a branch. Fixes src-d#313 Signed-off-by: Jeremy Chambers <jeremy@thehipbot.com> * dotgit: ignore filenames that don't match a hash For both packfiles and object files. Issue: keybase/client#11366 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * storage: dotgit, init fixtures in benchmark. Fixes src-d#770 fixtures is not initialized in BenchmarkRefMultipleTimes and caused panic. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: remote, Add shallow commits instead of substituting. Fixes src-d#412 updateShallow substituted the previous shallow list with the one returned by the UploadPackResponse. If the repository had previous shallow commits these are deleted from the list. This change adds the new shallow hashes to the old ones. Signed-off-by: Javi Fontan <jfontan@gmail.com> * dotgit: add test for bad file in pack directory Suggested by mcuadros. Issue: src-d#807 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * Resolve full commit sha to plumbing hash Signed-off-by: antham <hamonanth@gmail.com> * storage: filesystem, close shallow file when read Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * git: worktree, Skip special git directory. Fixes src-d#814 Signed-off-by: kuba-- <kuba@sourced.tech> * travis: dropping 1.8.x support due to golang.org/x/crypto/ssh requirement * Use remote name in fetch while clone Fixes src-d#827 Signed-off-by: Dustin Frisch <fooker@lab.sh> * Worktree: Provide ability to add excludes (src-d#825) Worktree: Provide ability to add excludes * Teach ResolveRevision how to look up annotated tags Signed-off-by: Mike Lundy <mike@fluffypenguin.org> * git: remote, Do not iterate all references on update. The current code iterates all the references in the remote to check if they match the refspec. This is OK when the refspec is a wildcard but is a waste of time when they are not. A hash with references is generated for fast access before starting the update and used only when the refspec is not a wildcard. In a repository with 7800 references this meant 7800 * 7800 checks. With the current code it took 8m30s to update the references. With the new code it takes less than 0.5s. References are already extensively tested in remote_test.go. Signed-off-by: Javi Fontan <jfontan@gmail.com> * idxfile: optimise allocations in readObjectNames This makes all the required Entry allocations in one go, instead of huge amounts of small individual allocations. Signed-off-by: David Symonds <dsymonds@golang.org> * packfile: improve Index memory representation to be more compact Instead of using a map for offset indexing, use a sorted slice. Binary searching is fast, and a slice is much more compact. This has a negligible hit on speed, but has a significant impact on memory usage, especially for larger repos. benchmark old ns/op new ns/op delta BenchmarkIndexConstruction-12 15506506 14056098 -9.35% benchmark old allocs new allocs delta BenchmarkIndexConstruction-12 60764 60385 -0.62% benchmark old bytes new bytes delta BenchmarkIndexConstruction-12 4318145 3913169 -9.38% Signed-off-by: David Symonds <dsymonds@golang.org> * config: modules, Ignore submodules with dotdot '..' path components. Fixes CVE-2018-11235 References: * https://blogs.msdn.microsoft.com/devops/2018/05/29/announcing-the-may-2018-git-security-vulnerability/ * https://security-tracker.debian.org/tracker/CVE-2018-11235 * git/git@0383bbb Signed-off-by: Joseph Vusich <jvusich@amazon.com> * worktree: Don't allow .gitmodules to be a symlink. Fixes CVE-2018-11235 References: * https://blogs.msdn.microsoft.com/devops/2018/05/29/announcing-the-may-2018-git-security-vulnerability/ * https://security-tracker.debian.org/tracker/CVE-2018-11235 * git/git@10ecfa7 Signed-off-by: Joseph Vusich <jvusich@amazon.com> * dotgit: Move package outside internal. Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * Remove println Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * plumbing: object, adds tree path cache to trees. Fixes src-d#793 The cache is used in Tree.FindEntry for faster path search. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: packfile, Don't push empty objects. Fixes src-d#840 Signed-off-by: kuba-- <kuba@sourced.tech> * storage: filesystem, make ObjectStorage constructor public Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing/transport: http, Adds token authentication support [Fixes src-d#858] Signed-off-by: Eric Billingsley <ebilling@babrains.com> * Fix documentation for Notes It previously said that it returned all references that are branches, but that's not true. Signed-off-by: Morgan Bazalgette <the@howl.moe> * packfile: optimise NewIndexFromIdxFile for a very common case Loading from an on-disk idxfile will usually already have the idxfile entries in order, so check that before wasting time on sorting. Signed-off-by: David Symonds <dsymonds@golang.org> * Remote.Fetch: error on missing remote reference Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * storage/filesystem: avoid norwfs build flag norwfs build flag was used to work on filesystems that do not support neither opening a file in read/write mode or renaming a file (e.f. sivafs). This had two problems: - go-git could not be compiled to work properly both with regular filesystems and limited filesystems at the same time. - the norwfs trick was not available on Windows. This PR removes the norwfs build flag, as well as the windows conditional flag on the dotgit package. For the file open mode, we use the new billy capabilities, to check at runtime if the filesystem supports opening a file in read/write mode or not. For the renaming, we just try and fallback to alternative methods if billy.ErrNotSupported is returned. Signed-off-by: Santiago M. Mola <santi@mola.io> * utils: diff, skip useless rune->string conversion According to library documentation : https://github.com/sergi/go-diff/blob/master/diffmatchpatch/diff.go#L391 Signed-off-by: Marc Barussaud <marc.barussaud@orange.com> * plumbing: add context to allow cancel on diff/patch computing Signed-off-by: Marc Barussaud <marc.barussaud@orange.com> * worktree: add test for correct tree sorting (issue src-d#881) Signed-off-by: Mark Bartel <github@spottybenny.ca> * worktree: sort the tree object. Fixes src-d#881 Signed-off-by: Mark Bartel <github@spottybenny.ca> * worktree: address PR comments: sort imports appropriately Signed-off-by: Mark Bartel <github@spottybenny.ca> * plumbing: object, expose ErrEntryNotFound in FindEntry. Fixes src-d#883 FindEntry will return ErrDirNotFound if the directory doesn't exist. But it doesn't return a public error if the entry itself is missing. This exposes the internal error ErrEntryNotFound, so users can programmatically check for this condition. Signed-off-by: James Ravn <james@r-vn.org> * plumbing/transport/internal: common, add support of Gogs for ErrRepositoryNotFound, avoiding to get an 'unknown error: '. Add some tests for existing supported services (github, gitlab, etc...) too. Signed-off-by: Jerome Doucet <jerdct@gmail.com> * plumbing/object: fix pgp signature encoder/decoder The way of reading pgp signatures was searching for pgp begin line in the header. This caused problems when this string appeared and was not part of the signature. For example if it appears in the message as an example or is part of the author name the decoder starts treating it as a signature. In this state the code was not able to notice then the header ended so it entered in an infinite loop searching for pgp end string. Now it uses the same method as original git. Searches for gpgsig section in header and starts getting all lines until the next part. In encoder the string used to add signatures was incorrect. It is now changed to the proper "gpgsig" string instead of "pgpsig". Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/format/idxfile: add new Index and MemoryIndex Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing/packfile: add new packfile parser Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: disable lookup by offset In one case it disables the cache and the other disables lookup when the scanner is not seekable. Could be added back later. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: idxfile, add idxfile.Writer with Observer interface It's still not complete: * 64 bit offsets * IdxChecksum Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: use Entry to hold object data Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: support offset64 generating indexes Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: preallocate memory in PatchDelta Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: fix bug searching in MemoryIndex Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: add offset/hash mapping to index This functionality may be moved elsewhere in the future but is needed now to fit filesystem.ObjectStorage and the new index. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: index is created only once and retrieved with Index Index is also automatically generated when OnFooter is called. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing, storage: integrate new index Now dotgit.PackWriter uses the new packfile.Parser and index. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: packfile, new Packfile representation Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing, packfile: delete index_test as is no longer used Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: fix two errors in idxfile and packfile decoder Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: add back IndexStorage Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: packfile, lazy object reads with DiskObjects Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing/idxfile: test FindHash and writer with 64 bit offsets Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: remove duplicated IndexStorage Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: add index generation to decoder Signed-off-by: Javi Fontan <jfontan@gmail.com> * Fix wrong godoc on Tags() method. Reword Tags() method documentation. Point to TagObjects() method to get all the tags on a repository. Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * plumbing: packfile, fix package tests Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * *: use parser to populate non writable storages and bug fixes Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * Fixed cloning of a single tag Relates to src-d#870 Signed-off-by: Fedor Korotkov <fedor.korotkov@gmail.com> * plumbing: packfile, allow non-seekable sources on Parser Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, add Parse benchmark Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, read object content only once Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * storage: filesystem, benchmark PackfileIter Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, rename DiskObject to FSObject Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * storage: filesystem, close Packfile after iterating objects Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * storage: filesystem, add PackfileIter benchmark reading object content Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, open and close packfile on FSObject reads Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * git: add benchmark for iterating repository objects Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: idxfile, Crc32 to CRC32 and return ok from findHashIndex Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: add buffer cache and use it in packfile parser It uses less memory and is faster as slices don't have to be converted from/to MemoryObject and they are indexed by offset. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/pacfile: tidy up objectInfo struct * a new hasher is created when needed * delete unused fields * base content is no longer kept in memory Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: do not compute sha1 for already undeltified objects Signed-off-by: Javi Fontan <jfontan@gmail.com> * added hook support Signed-off-by: noxora <ldecker@indeed.com> trying a possible fix to the delete test Signed-off-by: noxora <ldecker@indeed.com> still trying to fix this test Signed-off-by: noxora <ldecker@indeed.com> fixes did not work, seems to be a windows env problem Signed-off-by: noxora <ldecker@indeed.com> * plumbing: object, Don't add new line at end of commit signature The way that commit signatures were being written out was causing an extra newline to be written at the end of the commit when the message encoding was already taking care of this. Ultimately, this results in a corrupt object, rendering the object unverifiable with the signature in the commit. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add ability to PGP sign commits This adds the ability to sign commits by adding the SignKey field to CommitOptions. If present, the commit will be signed during the WorkTree.Commit call. The supplied SignKey must already be decrypted by the caller. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Remove use of strings.Builder This was added in Go 1.10 and is not supported on Go 1.9. Switched to bytes.Buffer to ensure compatibility. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Remove old hash validation code This will not work for a signed commit as with the GPG signature being a part of the commit, the hash is now non-deterministic. Verification of the commit is done through the validation of the signature. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add extra test for testing bad key error case I'm hoping this helps get codecov to a tolerable delta. :) Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * dotgit: fix object delete test Signed-off-by: Santiago M. Mola <santi@mola.io> * object: fix panic when reading object header When the first line of the pgp signature is an empty line or some header is malformed it crashes as there's no data for the header element. For example, if author name is "\n". Signed-off-by: Javi Fontan <jfontan@gmail.com> * Fixed an edge case for .gitignore Fixes src-d#923 Signed-off-by: Fedor Korotkov <fedor.korotkov@gmail.com> * plumbing/idxfile: object iterators returns entries in offset order In the latest change the order was changed from offset order in packfiles to hash order. This makes reading all the objects not as efficient as before. It also created problems when the previous order was expected. Also added EntriesByOffset to indexes. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: Add tagging support This adds a few methods: * CreateTag, which can be used to create both lightweight and annotated tags with a supplied TagObjectOptions struct. PGP signing is possible as well. * Tag, to fetch a single tag ref. As opposed to Tags or TagObjects, this will also fetch the tag object if it exists and return it along with the output. Lightweight tags just return the object as nil. * DeleteTag, to delete a tag. This simply deletes the ref. The object is left orphaned to be GCed later. I'm not 100% sure if DeleteTag is the correct behavior - looking for details on exactly *what* happens to a tag object if you delete the ref and not the tag were sparse, and groking the Git source did not really produce much insight to the untrained eye. This may be something that comes up in review. If deletion of the object is necessary, the in-memory storer may require some updates to allow DeleteLooseObject to be supported. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * plumbing: object, correct tag PGP encoding As with the update in ec3d2a8, tag encoding needed to be corrected to ensure extra newlines were not being added in during tag object encoding, so that it did not corrupt the object for verification. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * plumbing: object, don't add extra newline on PGP signature Tag encoding/decoding seems to be a lot more sensitive to requiring the exact expected format in the object, which generally includes messages canonicalized so that they have a newline on the end (even if they didn't before). As such, the message should be written with the newline (no need for an extra), and the PGP signature right after that, which will be newline split already, so there's no need to split it again. All of this means it's very important for the caller to send the message in the correct format - which I'm correcting in the next commit. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Canonicalize incoming annotated tag messages Tag messages are highly sensitive to being in the expected format, especially when encoding/decoding for PGP verification. As such, we do a simple trimming of whitespace on the incoming message and add a newline on the end, to ensure there are no surprises here. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Replace test signing key with one with longer expiry The old one was created with defaults, which would have caused CI failures in 2 years. The new one is valid for 10 years: > gpg --list-secret-keys /root/.gnupg/pubring.kbx ------------------------ sec rsa4096 2018-08-22 [SC] [expires: 2028-08-19] 93A17FF01E54328546087C8E029395402EFCCD53 uid [ unknown] foo bar <foo@foo.foo> Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * plumbing, storage: add bases to the common cache After clone only resolved deltas were added to the cache. This caused slowdowns in small repositories where most objects can be held in cache. It also makes packfiles reuse delta cache from the store. Previously it created a new delta cache each time a packfile object was created. This also slowed down a bit accessing objects and had an impact on memory consumption when bases are added to the cache. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: Discern tag target type from supplied hash I figured there was a way to do this without having to have TagObjectOptions supply this in - there is. Added support for this in and removed the object type from TagObjectOptions. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Don't return tag object with Tag, adjust docs for Tag and Tags I've mainly noticed that in using the current Tag function, that there were lots of times that I was ignoring the ref or the object, depending on what I needed. This was evident in the tests as well. As such, I think it just makes more sense for a singular tag fetcher to return just a ref, through which the caller may grab the annotation if they need it, and if it exists. Also, contrary to the docs on Tags, all tags have a ref, even if they are annotated. The difference between a lightweight tag and an annotated tag is the presence of the tag object, which the ref will point to if the tag is annotated. As such I've adjusted the docs with an example as to how one can get the annotation for a tag ref through the iterator. Source: https://git-scm.com/book/en/v2/Git-Internals-Git-References, tags section. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * storage/dotgit: search for incoming dir only once Search for incoming object directory was done once each time objects were accessed. This means a ReadDir of the objects path that is expensive. Now incoming directory is searched the first time an object is accessed and its name kept in DotGit to be reused. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/dotgit: use HasPrefix instead of Split Also reformatted function comment and fixed some typos. Signed-off-by: Javi Fontan <jfontan@gmail.com> * Remove empty dirs when cleaning with Dir opt. Signed-off-by: kuba-- <kuba@sourced.tech> * Add Status.IsUntracked function Signed-off-by: kuba-- <kuba@sourced.tech> * plumbing: object: Clamp object timestamps before unix epoch to unix epoch Signed-off-by: Taru Karttunen <taruti@taruti.net> * config: add commentChar to core config struct Signed-off-by: Zaq? Wiedmann <zaquestion@gmail.com> * git: add Static option to PlainOpen Also adds Static configuration to Storage and DotGit. This option means that the git repository is not expected to be modified while open and enables some optimizations. Each time a file is accessed the storer tries to open an object file for the requested hash. When this is done for a lot of objects it is expensive. With Static option a list of object files is generated the first time an object is accessed and used to check if exists instead of using system calls. A similar optimization is done for packfiles. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git, storer: use a common storer.Options for storer and PlainOpen Signed-off-by: Javi Fontan <jfontan@gmail.com> * dotgit: fix typo in comment Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: do not expose storage options in PlainOpen Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/storer: rename Static option to ExclusiveAccess Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: make Storage options private Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: move Options to filesytem and dotgit Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/dotgit: add ExclusiveAccess tests in dotgit This functionality was already tested in storage/filesystem. The coverage tool only takes into account files from the same package of the test. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/dotgit: add KeepDescriptors option This option maintains packfile file descriptors opened after reading objects from them. It improves performance as it does not have to be opening packfiles each time an object is needed. Also adds Close to EncodedObjectStorer to close all the files manualy. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/storer: do not expose Close in EncodedObjectStorer interface Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: add KeepDescriptors test Also delete Close from MockObjectStorage and memory storer. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: compare files using offset in test Using equals to compare files it uses diff to do so. This can potentially consume lots of ram. Changed the comparison to use file offsets. If the descriptor is reused the offset is maintained. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/transport: ssh check if list of known_hosts files is empty Signed-off-by: kuba-- <kuba@sourced.tech> * Fix fatal corrupt patch in unified diff format Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * Expose Storage cache. Signed-off-by: kuba-- <kuba@sourced.tech> * git: s/fetch/returns/ on Tag function doc This is to avoid any ambiguity with the act of "fetching" in git in general. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add Tag objects to the list of supported objects for walking This is necessary to support pruning on Tag objects. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Don't touch tag objects orphaned by tag deletion Deleting a tag ref for an annotated tag in normal git behavior does not delete the tag object right away. This is handled by the normal GC process. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add some tests for annotated tag deletion Added a couple of tests for annotated tag deletion: * The first one is a general test and should work regardless of the fixture used - the tag object could possibly be packed, so we do a prune *and* a repack operation before testing to see if the object was GCed correctly. * The second one actually creates the tag to be deleted, so that the tag object gets created as a loose, unpacked object. This is so we can effectively test that purning unpacked objects is now working 100% correctly (this was failing before because tag objects were not supported for walking). Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: s/TagObjectOptions/CreateTagOptions/ Just renaming the TagObjectOptions type to CreateTagOptions so that it's consistent with the other option types. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * repository: fix test for new Storage constructor Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * *: go modules support Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * travis: drop go1.9 add go1.11 * Fix potential LRU cache size issue. Signed-off-by: kuba-- <kuba@sourced.tech> * Remove empty space to trigger windows build. Signed-off-by: kuba-- <kuba@sourced.tech> * storage/filesystem: keep packs open in PackfileIter PackfileIter was not taking into account the option KeepDescriptors and was always closing the file. This caused "file already closed" errors when iterating packfiles in with KeepDescriptors active. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: add more doc to NewPackfileIter Signed-off-by: Javi Fontan <jfontan@gmail.com> * all: remove extra 's' in "mismatch" Signed-off-by: Jongmin Kim <jmkim@pukyong.ac.kr> * test: improve test for urlencoded user:pass Signed-off-by: Santiago M. Mola <santi@mola.io> * use time.IsZero in Prune Signed-off-by: u5surf <u5.horie@gmail.com> * Add test for Windows local paths. Signed-off-by: Filip Navara <navara@emclient.com> * git: Fix Status.IsClean() documentation The documentation of the IsClean Method contained a negation, so it was describing the opposite of its actual behavior. Fixes src-d#838 Signed-off-by: David Url <david@urld.io> * Plumbing: object, Add support for Log with filenames. Fixes src-d#826 (src-d#979) plumbing: object, Add support for Log with filenames. Fixes src-d#826 * object: get object size without reading whole object Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * tree: add a Size() method for getting plaintext size Without reading the entire object into memory. Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * filesystem: add a new test for EncodedObjectSize Suggested by taruti. Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * repository: allow open non-bare repositories as bare Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * use remote name in fetch while clone, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * references: sort: compare author timestamps when commit timestamps are equal, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * teach ResolveRevision how to look up annotated tags, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * teach ResolveRevision how to look up annotated tags, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * packfile: add comment on GetSizeByOffset Suggested by mcuadros. Issue: src-d#982 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * blame: fix edge case with missing \n in content length causing mismatched length error Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * repository: improve CheckoutOption.Hash doc Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * remote: use reference deltas on push when the remote server does not support offset deltas Signed-off-by: Benjamin Ash <bash@intelerad.com> * Fixed a typo. (src-d#989) README: Fixed a typo. * Enables building on openbsd, dragonfly bsd and solaris Signed-off-by: Yuce Tekol <yucetekol@gmail.com> * plumbing/format/packfile: Fix broken "thin" packfile support. Fixes src-d#991 Signed-off-by: Javier Peletier <jm@epiclabs.io> * plumbing: ReferenceName constructors Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * examples & documentation: PlainClone with Basic Authentication (Password & Access Token) (src-d#990) examples: PlainClone with Basic Authentication (Password & Access Token) * add StackOverflow to support channels Since we are not redirecting users to StackOverflow for support questions, it makes sense to add it to the official support channels. Signed-off-by: Santiago M. Mola <santi@mola.io> * plumbing: transport/http, Add missing host/port on redirect. Fixes src-d#820 Signed-off-by: Dave Henderson <dhenderson@gmail.com> * Fix spelling and grammar in docs and example Signed-off-by: Lukasz Kokot <lukasz@kumojin.com> * update gcfg dependency to v1.4.0 Signed-off-by: Dave Henderson <dhenderson@gmail.com> * repository: added cleanup for the PlainCloneContext() Signed-off-by: Bartek Jaroszewski <jaroszewskibartek@gmail.com> * improve cleanup implementation, add more tests Signed-off-by: Santiago M. Mola <santi@mola.io> * Update LICENSE Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * http: improve TokenAuth documentation Users are often confused with TokenAuth, since it might look that it should be used with GitHub's OAuth tokens. But that is not the case. TokenAuth implements HTTP bearer authentication. Most git servers will use HTTP basic authentication (user+passwords) even for OAuth tokens. Signed-off-by: Santiago M. Mola <santi@mola.io> * plumbing: ssh, Fix flaky test TestAdvertisedReferencesNotExists. Fixes src-d#969 Signed-off-by: Colton McCurdy <mccurdyc22@gmail.com> * repository: Fix RefSpec for a single tag. Fixes src-d#960 Signed-off-by: Fedor Korotkov <fedor.korotkov@gmail.com> * storage/filesystem: Added reindex method to reindex packfiles Signed-off-by: Javier Peletier <jm@epiclabs.io> * plumbing/format/packfile: Added thin pack test Signed-off-by: Javier Peletier <jm@epiclabs.io> * Remove unused method (src-d#1022) Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * plumbing: format/index: support for EOIE extension, by default on git v2.2.0 Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * repository: fix plain clone error handling regression PR src-d#1008 introduced a regression by changing the errors returned by PlainClone when a repository did not exist. This change goes back to returned errors as they were in v4.7.0. Fixes src-d#1027 Signed-off-by: Santiago M. Mola <santi@mola.io> * plumbing: format/packfile, performance optimizations for reading large commit histories (src-d#963) Signed-off-by: Filip Navara <navara@emclient.com>
Snehal1112
added a commit
to Snehal1112/go-git
that referenced
this pull request
Dec 2, 2018
* plumbing: format: pktline, Accept oversized pkt-lines up to 65524 bytes The canonical Git client successfully decodes sideband packets up to 65524 bytes in length (4-byte header + 65520-byte payload). The Git protocol documentation was updated in August 2016 to reduce the maximum payload size to 65516 bytes, however old implementations still exist in the wild emitting 65520-byte payloads. As there is no technical difficulty with accepting (not emitting) larger payload sizes, this change adjusts the limit check to allow successful decoding of packets up to 65524 bytes. This change increases compatibility with the current canonical Git implementation. Doc changes from August 2016: git/git@7841c48#diff-52695c8fe91b78b70cea44562ae28297L67 Current packet buffer size is still LARGE_PACKET_MAX (+1 null): https://github.com/git/git/blob/468165c1d8a442994a825f3684528361727cd8c0/sideband.c#L24 https://github.com/git/git/blob/468165c1d8a442994a825f3684528361727cd8c0/sideband.c#L36 LARGE_PACKET_MAX definition: https://github.com/git/git/blob/468165c1d8a442994a825f3684528361727cd8c0/pkt-line.h#L100 Signed-off-by: Joseph Vusich <jvusich@amazon.com> * add PlainOpen variant to find .git in parent dirs This is the git tool's behavior that people are used to; if one runs a git command in a repository's subdirectory, git still works. Fixes src-d#765. Signed-off-by: Daniel Martí <mvdan@mvdan.cc> * use bsd superset for conditional compilation Signed-off-by: wardn <wardn@users.noreply.github.com> * config: adds branches to config for tracking branches against remotes, updates clone to track when cloning a branch. Fixes src-d#313 Signed-off-by: Jeremy Chambers <jeremy@thehipbot.com> * dotgit: ignore filenames that don't match a hash For both packfiles and object files. Issue: keybase/client#11366 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * storage: dotgit, init fixtures in benchmark. Fixes src-d#770 fixtures is not initialized in BenchmarkRefMultipleTimes and caused panic. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: remote, Add shallow commits instead of substituting. Fixes src-d#412 updateShallow substituted the previous shallow list with the one returned by the UploadPackResponse. If the repository had previous shallow commits these are deleted from the list. This change adds the new shallow hashes to the old ones. Signed-off-by: Javi Fontan <jfontan@gmail.com> * dotgit: add test for bad file in pack directory Suggested by mcuadros. Issue: src-d#807 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * Resolve full commit sha to plumbing hash Signed-off-by: antham <hamonanth@gmail.com> * storage: filesystem, close shallow file when read Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * git: worktree, Skip special git directory. Fixes src-d#814 Signed-off-by: kuba-- <kuba@sourced.tech> * travis: dropping 1.8.x support due to golang.org/x/crypto/ssh requirement * Use remote name in fetch while clone Fixes src-d#827 Signed-off-by: Dustin Frisch <fooker@lab.sh> * Worktree: Provide ability to add excludes (src-d#825) Worktree: Provide ability to add excludes * Teach ResolveRevision how to look up annotated tags Signed-off-by: Mike Lundy <mike@fluffypenguin.org> * git: remote, Do not iterate all references on update. The current code iterates all the references in the remote to check if they match the refspec. This is OK when the refspec is a wildcard but is a waste of time when they are not. A hash with references is generated for fast access before starting the update and used only when the refspec is not a wildcard. In a repository with 7800 references this meant 7800 * 7800 checks. With the current code it took 8m30s to update the references. With the new code it takes less than 0.5s. References are already extensively tested in remote_test.go. Signed-off-by: Javi Fontan <jfontan@gmail.com> * idxfile: optimise allocations in readObjectNames This makes all the required Entry allocations in one go, instead of huge amounts of small individual allocations. Signed-off-by: David Symonds <dsymonds@golang.org> * packfile: improve Index memory representation to be more compact Instead of using a map for offset indexing, use a sorted slice. Binary searching is fast, and a slice is much more compact. This has a negligible hit on speed, but has a significant impact on memory usage, especially for larger repos. benchmark old ns/op new ns/op delta BenchmarkIndexConstruction-12 15506506 14056098 -9.35% benchmark old allocs new allocs delta BenchmarkIndexConstruction-12 60764 60385 -0.62% benchmark old bytes new bytes delta BenchmarkIndexConstruction-12 4318145 3913169 -9.38% Signed-off-by: David Symonds <dsymonds@golang.org> * config: modules, Ignore submodules with dotdot '..' path components. Fixes CVE-2018-11235 References: * https://blogs.msdn.microsoft.com/devops/2018/05/29/announcing-the-may-2018-git-security-vulnerability/ * https://security-tracker.debian.org/tracker/CVE-2018-11235 * git/git@0383bbb Signed-off-by: Joseph Vusich <jvusich@amazon.com> * worktree: Don't allow .gitmodules to be a symlink. Fixes CVE-2018-11235 References: * https://blogs.msdn.microsoft.com/devops/2018/05/29/announcing-the-may-2018-git-security-vulnerability/ * https://security-tracker.debian.org/tracker/CVE-2018-11235 * git/git@10ecfa7 Signed-off-by: Joseph Vusich <jvusich@amazon.com> * dotgit: Move package outside internal. Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * Remove println Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * plumbing: object, adds tree path cache to trees. Fixes src-d#793 The cache is used in Tree.FindEntry for faster path search. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: packfile, Don't push empty objects. Fixes src-d#840 Signed-off-by: kuba-- <kuba@sourced.tech> * storage: filesystem, make ObjectStorage constructor public Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing/transport: http, Adds token authentication support [Fixes src-d#858] Signed-off-by: Eric Billingsley <ebilling@babrains.com> * Fix documentation for Notes It previously said that it returned all references that are branches, but that's not true. Signed-off-by: Morgan Bazalgette <the@howl.moe> * packfile: optimise NewIndexFromIdxFile for a very common case Loading from an on-disk idxfile will usually already have the idxfile entries in order, so check that before wasting time on sorting. Signed-off-by: David Symonds <dsymonds@golang.org> * Remote.Fetch: error on missing remote reference Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * storage/filesystem: avoid norwfs build flag norwfs build flag was used to work on filesystems that do not support neither opening a file in read/write mode or renaming a file (e.f. sivafs). This had two problems: - go-git could not be compiled to work properly both with regular filesystems and limited filesystems at the same time. - the norwfs trick was not available on Windows. This PR removes the norwfs build flag, as well as the windows conditional flag on the dotgit package. For the file open mode, we use the new billy capabilities, to check at runtime if the filesystem supports opening a file in read/write mode or not. For the renaming, we just try and fallback to alternative methods if billy.ErrNotSupported is returned. Signed-off-by: Santiago M. Mola <santi@mola.io> * utils: diff, skip useless rune->string conversion According to library documentation : https://github.com/sergi/go-diff/blob/master/diffmatchpatch/diff.go#L391 Signed-off-by: Marc Barussaud <marc.barussaud@orange.com> * plumbing: add context to allow cancel on diff/patch computing Signed-off-by: Marc Barussaud <marc.barussaud@orange.com> * worktree: add test for correct tree sorting (issue src-d#881) Signed-off-by: Mark Bartel <github@spottybenny.ca> * worktree: sort the tree object. Fixes src-d#881 Signed-off-by: Mark Bartel <github@spottybenny.ca> * worktree: address PR comments: sort imports appropriately Signed-off-by: Mark Bartel <github@spottybenny.ca> * plumbing: object, expose ErrEntryNotFound in FindEntry. Fixes src-d#883 FindEntry will return ErrDirNotFound if the directory doesn't exist. But it doesn't return a public error if the entry itself is missing. This exposes the internal error ErrEntryNotFound, so users can programmatically check for this condition. Signed-off-by: James Ravn <james@r-vn.org> * plumbing/transport/internal: common, add support of Gogs for ErrRepositoryNotFound, avoiding to get an 'unknown error: '. Add some tests for existing supported services (github, gitlab, etc...) too. Signed-off-by: Jerome Doucet <jerdct@gmail.com> * plumbing/object: fix pgp signature encoder/decoder The way of reading pgp signatures was searching for pgp begin line in the header. This caused problems when this string appeared and was not part of the signature. For example if it appears in the message as an example or is part of the author name the decoder starts treating it as a signature. In this state the code was not able to notice then the header ended so it entered in an infinite loop searching for pgp end string. Now it uses the same method as original git. Searches for gpgsig section in header and starts getting all lines until the next part. In encoder the string used to add signatures was incorrect. It is now changed to the proper "gpgsig" string instead of "pgpsig". Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/format/idxfile: add new Index and MemoryIndex Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing/packfile: add new packfile parser Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: disable lookup by offset In one case it disables the cache and the other disables lookup when the scanner is not seekable. Could be added back later. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: idxfile, add idxfile.Writer with Observer interface It's still not complete: * 64 bit offsets * IdxChecksum Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: use Entry to hold object data Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: support offset64 generating indexes Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: preallocate memory in PatchDelta Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: fix bug searching in MemoryIndex Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: add offset/hash mapping to index This functionality may be moved elsewhere in the future but is needed now to fit filesystem.ObjectStorage and the new index. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/idxfile: index is created only once and retrieved with Index Index is also automatically generated when OnFooter is called. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing, storage: integrate new index Now dotgit.PackWriter uses the new packfile.Parser and index. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: packfile, new Packfile representation Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing, packfile: delete index_test as is no longer used Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: fix two errors in idxfile and packfile decoder Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: add back IndexStorage Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing: packfile, lazy object reads with DiskObjects Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing/idxfile: test FindHash and writer with 64 bit offsets Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: remove duplicated IndexStorage Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: add index generation to decoder Signed-off-by: Javi Fontan <jfontan@gmail.com> * Fix wrong godoc on Tags() method. Reword Tags() method documentation. Point to TagObjects() method to get all the tags on a repository. Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * plumbing: packfile, fix package tests Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * *: use parser to populate non writable storages and bug fixes Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * Fixed cloning of a single tag Relates to src-d#870 Signed-off-by: Fedor Korotkov <fedor.korotkov@gmail.com> * plumbing: packfile, allow non-seekable sources on Parser Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, add Parse benchmark Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, read object content only once Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * storage: filesystem, benchmark PackfileIter Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, rename DiskObject to FSObject Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * storage: filesystem, close Packfile after iterating objects Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * storage: filesystem, add PackfileIter benchmark reading object content Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: packfile, open and close packfile on FSObject reads Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * git: add benchmark for iterating repository objects Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: idxfile, Crc32 to CRC32 and return ok from findHashIndex Signed-off-by: Miguel Molina <miguel@erizocosmi.co> * plumbing: add buffer cache and use it in packfile parser It uses less memory and is faster as slices don't have to be converted from/to MemoryObject and they are indexed by offset. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/pacfile: tidy up objectInfo struct * a new hasher is created when needed * delete unused fields * base content is no longer kept in memory Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/packfile: do not compute sha1 for already undeltified objects Signed-off-by: Javi Fontan <jfontan@gmail.com> * added hook support Signed-off-by: noxora <ldecker@indeed.com> trying a possible fix to the delete test Signed-off-by: noxora <ldecker@indeed.com> still trying to fix this test Signed-off-by: noxora <ldecker@indeed.com> fixes did not work, seems to be a windows env problem Signed-off-by: noxora <ldecker@indeed.com> * plumbing: object, Don't add new line at end of commit signature The way that commit signatures were being written out was causing an extra newline to be written at the end of the commit when the message encoding was already taking care of this. Ultimately, this results in a corrupt object, rendering the object unverifiable with the signature in the commit. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add ability to PGP sign commits This adds the ability to sign commits by adding the SignKey field to CommitOptions. If present, the commit will be signed during the WorkTree.Commit call. The supplied SignKey must already be decrypted by the caller. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Remove use of strings.Builder This was added in Go 1.10 and is not supported on Go 1.9. Switched to bytes.Buffer to ensure compatibility. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Remove old hash validation code This will not work for a signed commit as with the GPG signature being a part of the commit, the hash is now non-deterministic. Verification of the commit is done through the validation of the signature. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add extra test for testing bad key error case I'm hoping this helps get codecov to a tolerable delta. :) Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * dotgit: fix object delete test Signed-off-by: Santiago M. Mola <santi@mola.io> * object: fix panic when reading object header When the first line of the pgp signature is an empty line or some header is malformed it crashes as there's no data for the header element. For example, if author name is "\n". Signed-off-by: Javi Fontan <jfontan@gmail.com> * Fixed an edge case for .gitignore Fixes src-d#923 Signed-off-by: Fedor Korotkov <fedor.korotkov@gmail.com> * plumbing/idxfile: object iterators returns entries in offset order In the latest change the order was changed from offset order in packfiles to hash order. This makes reading all the objects not as efficient as before. It also created problems when the previous order was expected. Also added EntriesByOffset to indexes. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: Add tagging support This adds a few methods: * CreateTag, which can be used to create both lightweight and annotated tags with a supplied TagObjectOptions struct. PGP signing is possible as well. * Tag, to fetch a single tag ref. As opposed to Tags or TagObjects, this will also fetch the tag object if it exists and return it along with the output. Lightweight tags just return the object as nil. * DeleteTag, to delete a tag. This simply deletes the ref. The object is left orphaned to be GCed later. I'm not 100% sure if DeleteTag is the correct behavior - looking for details on exactly *what* happens to a tag object if you delete the ref and not the tag were sparse, and groking the Git source did not really produce much insight to the untrained eye. This may be something that comes up in review. If deletion of the object is necessary, the in-memory storer may require some updates to allow DeleteLooseObject to be supported. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * plumbing: object, correct tag PGP encoding As with the update in ec3d2a8, tag encoding needed to be corrected to ensure extra newlines were not being added in during tag object encoding, so that it did not corrupt the object for verification. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * plumbing: object, don't add extra newline on PGP signature Tag encoding/decoding seems to be a lot more sensitive to requiring the exact expected format in the object, which generally includes messages canonicalized so that they have a newline on the end (even if they didn't before). As such, the message should be written with the newline (no need for an extra), and the PGP signature right after that, which will be newline split already, so there's no need to split it again. All of this means it's very important for the caller to send the message in the correct format - which I'm correcting in the next commit. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Canonicalize incoming annotated tag messages Tag messages are highly sensitive to being in the expected format, especially when encoding/decoding for PGP verification. As such, we do a simple trimming of whitespace on the incoming message and add a newline on the end, to ensure there are no surprises here. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Replace test signing key with one with longer expiry The old one was created with defaults, which would have caused CI failures in 2 years. The new one is valid for 10 years: > gpg --list-secret-keys /root/.gnupg/pubring.kbx ------------------------ sec rsa4096 2018-08-22 [SC] [expires: 2028-08-19] 93A17FF01E54328546087C8E029395402EFCCD53 uid [ unknown] foo bar <foo@foo.foo> Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * plumbing, storage: add bases to the common cache After clone only resolved deltas were added to the cache. This caused slowdowns in small repositories where most objects can be held in cache. It also makes packfiles reuse delta cache from the store. Previously it created a new delta cache each time a packfile object was created. This also slowed down a bit accessing objects and had an impact on memory consumption when bases are added to the cache. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: Discern tag target type from supplied hash I figured there was a way to do this without having to have TagObjectOptions supply this in - there is. Added support for this in and removed the object type from TagObjectOptions. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Don't return tag object with Tag, adjust docs for Tag and Tags I've mainly noticed that in using the current Tag function, that there were lots of times that I was ignoring the ref or the object, depending on what I needed. This was evident in the tests as well. As such, I think it just makes more sense for a singular tag fetcher to return just a ref, through which the caller may grab the annotation if they need it, and if it exists. Also, contrary to the docs on Tags, all tags have a ref, even if they are annotated. The difference between a lightweight tag and an annotated tag is the presence of the tag object, which the ref will point to if the tag is annotated. As such I've adjusted the docs with an example as to how one can get the annotation for a tag ref through the iterator. Source: https://git-scm.com/book/en/v2/Git-Internals-Git-References, tags section. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * storage/dotgit: search for incoming dir only once Search for incoming object directory was done once each time objects were accessed. This means a ReadDir of the objects path that is expensive. Now incoming directory is searched the first time an object is accessed and its name kept in DotGit to be reused. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/dotgit: use HasPrefix instead of Split Also reformatted function comment and fixed some typos. Signed-off-by: Javi Fontan <jfontan@gmail.com> * Remove empty dirs when cleaning with Dir opt. Signed-off-by: kuba-- <kuba@sourced.tech> * Add Status.IsUntracked function Signed-off-by: kuba-- <kuba@sourced.tech> * plumbing: object: Clamp object timestamps before unix epoch to unix epoch Signed-off-by: Taru Karttunen <taruti@taruti.net> * config: add commentChar to core config struct Signed-off-by: Zaq? Wiedmann <zaquestion@gmail.com> * git: add Static option to PlainOpen Also adds Static configuration to Storage and DotGit. This option means that the git repository is not expected to be modified while open and enables some optimizations. Each time a file is accessed the storer tries to open an object file for the requested hash. When this is done for a lot of objects it is expensive. With Static option a list of object files is generated the first time an object is accessed and used to check if exists instead of using system calls. A similar optimization is done for packfiles. Signed-off-by: Javi Fontan <jfontan@gmail.com> * git, storer: use a common storer.Options for storer and PlainOpen Signed-off-by: Javi Fontan <jfontan@gmail.com> * dotgit: fix typo in comment Signed-off-by: Javi Fontan <jfontan@gmail.com> * git: do not expose storage options in PlainOpen Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/storer: rename Static option to ExclusiveAccess Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: make Storage options private Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: move Options to filesytem and dotgit Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/dotgit: add ExclusiveAccess tests in dotgit This functionality was already tested in storage/filesystem. The coverage tool only takes into account files from the same package of the test. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/dotgit: add KeepDescriptors option This option maintains packfile file descriptors opened after reading objects from them. It improves performance as it does not have to be opening packfiles each time an object is needed. Also adds Close to EncodedObjectStorer to close all the files manualy. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/storer: do not expose Close in EncodedObjectStorer interface Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: add KeepDescriptors test Also delete Close from MockObjectStorage and memory storer. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: compare files using offset in test Using equals to compare files it uses diff to do so. This can potentially consume lots of ram. Changed the comparison to use file offsets. If the descriptor is reused the offset is maintained. Signed-off-by: Javi Fontan <jfontan@gmail.com> * plumbing/transport: ssh check if list of known_hosts files is empty Signed-off-by: kuba-- <kuba@sourced.tech> * Fix fatal corrupt patch in unified diff format Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * Expose Storage cache. Signed-off-by: kuba-- <kuba@sourced.tech> * git: s/fetch/returns/ on Tag function doc This is to avoid any ambiguity with the act of "fetching" in git in general. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add Tag objects to the list of supported objects for walking This is necessary to support pruning on Tag objects. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Don't touch tag objects orphaned by tag deletion Deleting a tag ref for an annotated tag in normal git behavior does not delete the tag object right away. This is handled by the normal GC process. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: Add some tests for annotated tag deletion Added a couple of tests for annotated tag deletion: * The first one is a general test and should work regardless of the fixture used - the tag object could possibly be packed, so we do a prune *and* a repack operation before testing to see if the object was GCed correctly. * The second one actually creates the tag to be deleted, so that the tag object gets created as a loose, unpacked object. This is so we can effectively test that purning unpacked objects is now working 100% correctly (this was failing before because tag objects were not supported for walking). Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * git: s/TagObjectOptions/CreateTagOptions/ Just renaming the TagObjectOptions type to CreateTagOptions so that it's consistent with the other option types. Signed-off-by: Chris Marchesi <chrism@vancluevertech.com> * repository: fix test for new Storage constructor Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * *: go modules support Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * travis: drop go1.9 add go1.11 * Fix potential LRU cache size issue. Signed-off-by: kuba-- <kuba@sourced.tech> * Remove empty space to trigger windows build. Signed-off-by: kuba-- <kuba@sourced.tech> * storage/filesystem: keep packs open in PackfileIter PackfileIter was not taking into account the option KeepDescriptors and was always closing the file. This caused "file already closed" errors when iterating packfiles in with KeepDescriptors active. Signed-off-by: Javi Fontan <jfontan@gmail.com> * storage/filesystem: add more doc to NewPackfileIter Signed-off-by: Javi Fontan <jfontan@gmail.com> * all: remove extra 's' in "mismatch" Signed-off-by: Jongmin Kim <jmkim@pukyong.ac.kr> * test: improve test for urlencoded user:pass Signed-off-by: Santiago M. Mola <santi@mola.io> * use time.IsZero in Prune Signed-off-by: u5surf <u5.horie@gmail.com> * Add test for Windows local paths. Signed-off-by: Filip Navara <navara@emclient.com> * git: Fix Status.IsClean() documentation The documentation of the IsClean Method contained a negation, so it was describing the opposite of its actual behavior. Fixes src-d#838 Signed-off-by: David Url <david@urld.io> * Plumbing: object, Add support for Log with filenames. Fixes src-d#826 (src-d#979) plumbing: object, Add support for Log with filenames. Fixes src-d#826 * object: get object size without reading whole object Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * tree: add a Size() method for getting plaintext size Without reading the entire object into memory. Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * filesystem: add a new test for EncodedObjectSize Suggested by taruti. Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * repository: allow open non-bare repositories as bare Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * use remote name in fetch while clone, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * references: sort: compare author timestamps when commit timestamps are equal, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * teach ResolveRevision how to look up annotated tags, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * teach ResolveRevision how to look up annotated tags, test Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * packfile: add comment on GetSizeByOffset Suggested by mcuadros. Issue: src-d#982 Signed-off-by: Jeremy Stribling <strib@alum.mit.edu> * blame: fix edge case with missing \n in content length causing mismatched length error Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * repository: improve CheckoutOption.Hash doc Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * remote: use reference deltas on push when the remote server does not support offset deltas Signed-off-by: Benjamin Ash <bash@intelerad.com> * Fixed a typo. (src-d#989) README: Fixed a typo. * Enables building on openbsd, dragonfly bsd and solaris Signed-off-by: Yuce Tekol <yucetekol@gmail.com> * plumbing/format/packfile: Fix broken "thin" packfile support. Fixes src-d#991 Signed-off-by: Javier Peletier <jm@epiclabs.io> * plumbing: ReferenceName constructors Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * examples & documentation: PlainClone with Basic Authentication (Password & Access Token) (src-d#990) examples: PlainClone with Basic Authentication (Password & Access Token) * add StackOverflow to support channels Since we are not redirecting users to StackOverflow for support questions, it makes sense to add it to the official support channels. Signed-off-by: Santiago M. Mola <santi@mola.io> * plumbing: transport/http, Add missing host/port on redirect. Fixes src-d#820 Signed-off-by: Dave Henderson <dhenderson@gmail.com> * Fix spelling and grammar in docs and example Signed-off-by: Lukasz Kokot <lukasz@kumojin.com> * update gcfg dependency to v1.4.0 Signed-off-by: Dave Henderson <dhenderson@gmail.com> * repository: added cleanup for the PlainCloneContext() Signed-off-by: Bartek Jaroszewski <jaroszewskibartek@gmail.com> * improve cleanup implementation, add more tests Signed-off-by: Santiago M. Mola <santi@mola.io> * Update LICENSE Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * http: improve TokenAuth documentation Users are often confused with TokenAuth, since it might look that it should be used with GitHub's OAuth tokens. But that is not the case. TokenAuth implements HTTP bearer authentication. Most git servers will use HTTP basic authentication (user+passwords) even for OAuth tokens. Signed-off-by: Santiago M. Mola <santi@mola.io> * plumbing: ssh, Fix flaky test TestAdvertisedReferencesNotExists. Fixes src-d#969 Signed-off-by: Colton McCurdy <mccurdyc22@gmail.com> * repository: Fix RefSpec for a single tag. Fixes src-d#960 Signed-off-by: Fedor Korotkov <fedor.korotkov@gmail.com> * storage/filesystem: Added reindex method to reindex packfiles Signed-off-by: Javier Peletier <jm@epiclabs.io> * plumbing/format/packfile: Added thin pack test Signed-off-by: Javier Peletier <jm@epiclabs.io> * Remove unused method (src-d#1022) Signed-off-by: Antonio Jesus Navarro Perez <antnavper@gmail.com> * plumbing: format/index: support for EOIE extension, by default on git v2.2.0 Signed-off-by: Máximo Cuadros <mcuadros@gmail.com> * repository: fix plain clone error handling regression PR src-d#1008 introduced a regression by changing the errors returned by PlainClone when a repository did not exist. This change goes back to returned errors as they were in v4.7.0. Fixes src-d#1027 Signed-off-by: Santiago M. Mola <santi@mola.io> * plumbing: format/packfile, performance optimizations for reading large commit histories (src-d#963) Signed-off-by: Filip Navara <navara@emclient.com>
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For both packfiles and object files.
Issue: keybase/client#11366