Track down leaked ptp_getobjectinfo results
#690
Comments
|
i did a first pass. I however need to now test as many of those cameras as I can in case I made a mistake ;) |
|
Nice, thanks! Yeah, testing was my concern too - unfortunately, I don't have many cameras to verify it on. |
|
Since a few weeks i installed 2.5.27, and have also discovered a growing memory usage when running live previews. I have a machine vision system that is running non-stop and this memory leak will eventually lead to a fatal Out of Memory error. That system can't run for more than two days now... |
|
It's strange that this issue went unnoticed for such a long time though... as far as I can tell, most of the relevant code has been unchanged for years? |
Have you also looked for leaks associated with capture and download of images? My application does that a lot as well, if it sees something interesting in the preview. I will monitor memory usage and hopefully all the memory leaks are gone now! |
What is the magnitude of the leaked memory you were seeing? I am seeing on average 10.5 MB per hour, when capturing 25 previews a second. I measured the memory use of my application every 5 minutes, with the (ubuntu) system monitor, and the memory growth isn't linear. For example: All this time the camera is stationary, pointed at a scene that doesn't change, so the preview data is very similar in size on each new frame. There is a complication in my setup: i am programming in Go, and i use a wrapper around the C libgphoto2 library. It is possible that there is a flaw in that wrapper. Maybe it doesn't free everything that needs to be freed. I think my next experiment would be to write a pure C program that captures previews as fast as possible and see how the memory usage is, over longer periods of time. |
|
various live methods have only been added recently in the last 3 years or so FWIW. |
@notnot FWIW it's a very slow and unreliable method. Best to compile your code with leak sanitizers or run with Valgrind (I'm not entirely sure which one integrates with Go better). That's how I found the leak above, and ensured that at least my program didn't have any others, no matter how small. |
Yes this is a crude way of monitoring memory. I will learn about the best ways to do this accurately with Go executables. Go is a garbage collecting language, which is not so simple to reason about :( In the meantime i did write a minimal C program that captures previews 25 times a second, and i don't see any memory leaks when running it. The huge memory leak that i was seeing in my application must thus be on the Go side. I will investigate the gphoto2 wrapper in detail to find the flaw. Sorry for the noise! There is still a mystery though: with libgphoto2 2.5.20, which i have been using for some years, my application/machine could run for months without an OOM error. With 2.5.27 it runs for at most 2 days, then it hits an OOM. I didn't change my code, i only upgraded libgphoto2. |
|
I spoke too soon... when running the memory leak test for a longer time, i do see a serious leak related to preview captures. This happens in my C test program, so it has nothing to do with the Go wrapper. Note that the older 2.5.20 that i used to use in production doesn't suffer from it. 2.5.27 has it. I am running this test on arm cpu, maybe that matters. frame 346: 26343 bytes of data @ 0x55AFF6CFD0 In 2.5.20 the preview data is always stored at the same address, at least it is on another of my machines where i tested on too. No memory leaks there, even after an hour of previews at 25 FPS. It is possible that my C test program is not correct, i include it here, so that you can verify it:// gphoto2.c, jpad 2021 #include <stdio.h> #include <gphoto2/gphoto2.h> int main(int argc, char **argv) }If you need more info on my setup, please tell me how to provide it. My computer vision machine is running on a nVidia Jeston Nano computer. |
|
Note that i tried both gp_file_unref() and gp_file_free(). No difference. Or should i use both? |
Just to be clear - 2.5.27 release or the master of the repo (2.5.27.1)? Only the master has the fix for the leak. |
|
Of course i have applied the changes. The leak is serious, for example time mem 00.00 : 0.8 MB |
|
Can you build with |
|
i can try |
|
It will print it for you upon exit automatically. If you haven't used it before, the docs are pretty good: https://clang.llvm.org/docs/LeakSanitizer.html You can also reduce number of iterations - just run like 100 previews or so, if there's a leak, it doesn't matter how many you run, it should show up. |
Er, actually, that page doesn't show just |
The thing is... i think the leak comes later on... usually after several hundred previews.. see my earlier discovery of the way that the data location changes... probably related to that |
Maybe you just notice it at that point, but LeakSanitizer will tell you if you leak 4 bytes... |
Data location can change due to allocator fragmentation, it's not much of a symptom by itself usually. |
|
frame 90: 26413 bytes of data @ 0x7F7E700000 ================================================================= Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 4096 byte(s) in 1 object(s) allocated from: Direct leak of 3159 byte(s) in 243 object(s) allocated from: SUMMARY: LeakSanitizer: 35927 byte(s) leaked in 251 allocation(s). |
|
Nit: it's best to always wrap code, logs, etc. on Github with Otherwise you just accidentally cc'd a bunch of old issues. |
|
The above was just 100 preview frames |
|
@notnot You can pinpoint it quite easily yourself. Ignore everything coming from Unfortunately, looks like your libgphoto2 is built with optimisations and not just a debug build, so it doesn't provide much more info (missing function names). |
|
Make sure to rebuild everything - clean & make libgphoto2 as well as rebuild your own wrapper - with just |
Btw, if you free camera itself before exit, those spurious leaks should go away too. |
|
sorry... i'm quite new to this, second attempt with better formatting |
Will attempt that tomorrow, and i'll post the results. |
suggests there is still some code not rebuilt with debug symbols. But at least you can take a look at the line it gives at gphoto2-camera.c:1403 right before that. |
As I said above, I didn't see any more leaks in my app, but it could very well be camera-specific. |
I am running this with a Sony A6300, which uses ptp. I figure your A6600 will use the same code paths? |
|
Unfortunately, that line - 1403 - does not correspond to anything on current master, so it appears your copy is still out of date: libgphoto2/libgphoto2/gphoto2-camera.c Line 1403 in 04fed07 But, either way, mind you, I'm not a maintainer or anything - just a fellow user who happened to debug issues that arose with my camera and contributed some fixes and trying to help the best I can :) I'm sure you'll be able to find out what's causing leaks in your app too, you just need some time to go over those codepaths the leak indicated with a debugger or by inserting some logs. |
Yes that is true, i manually applied the changes in the official 2.5.27, because at the location where my machine is, the internet is very very bad. I was even lucky to get the diff page on screen! I should download the current master at home, save it to stick and install it on site. Tomorrow, or the day after... Thanks for your concerns, i hope Marcus will have a look at it. I lost most of my C skills lol. It feels like i entered the rabbit hole! Good thing is that i can now produce leak sanitizer logs, thanks to you :) |
|
Line 1403 in my gphoto2-camera-c source: |
Ah right, that's not very helpful then. A bit too high-level API to tell anything useful. You'll need to clean & recompile the rest with debug flags and without opts as described above to see missing frames and figure out where the leak actually occurs. |
|
the strdup in the leak trace could just be this ptp_getobjectinfo thing. build libgphoto2 also with CFLAGS="-O2 -DVALGRIND -g" .... the -DVALGRIND disables the unloading of the camera drivers so their symvbols stay in memory. |
|
I can confirm that today's master no longer has a preview capture memory leak. I still have two (not so important) questions:
|
Great, glad to hear! Sounds like it should've been came codepath then.
You might have missed some commits as the fixes were split between Sony and other cameras. That's why I asked earlier about
Judging by the fact those all are coming from |
|
it is probably just 1 time initialization, yes. without better debug symbols its a bit hard to say. (But I see similar one time leaks here.) |
|
I think it's fair to close this issue for now; testing-wise I assume people will report new issues when they run into problems. It could be really nice if gphoto tests could be automated somehow, but I doubt it's possible without some sort of full-grown camera emulators (or a server with a bunch of cameras connected somewhere). |
|
yes |
|
I wonder if it makes sense to at least build and ship gphoto2 CLI with Unlike full-on address sanitizer, it has very low overhead (definitely negligible compared to USB I/O speed as well as those (The false leaks above can be configured to be excluded so people won't see any leaks that are not real.) |
…on 2.5.28
Arkell Rasiah (2):
camlibs/ptp2/ptp-pack.c: Fix get-config aperture for R5.
camlibs/ptp2/ptp-pack.c: Qualify previous change only for olcver=0x13 cameras.
Daniel Schulte (20):
camlibs/ptp2: Always return after PTP/IP IO errors (#595)
WIP: Make PTP/IP implementation work on Windows (#596)
gp_camera_exit: Return result of camera exit function if possible
camlibs/ptp2: Better cross-platform error handling
camlibs/ptp2: fuji: Also close jpg socket if connection fails
camlibs/ptp2: Better network timeout handling
camlibs/ptp2: Add more error handling to camera_exit
camlibs/ptp2: Add timeout support for connect/read/write socket operations
camlibs/ptp2: Reduce log level of debug message
camlibs/ptp2: Unpack Canon EOS changes for INT32 properties
camlibs/ptp2: fuji: Fix property datatype in camera_trigger_capture
camlibs/ptp2: fuji: Ignore added preview object while live-view is active
camlibs/ptp2: Add Fuji priority mode setting
camlibs/ptp2: fuji: Use PriorityMode define in camera (un)prepare code
camlibs/ptp2: Fuji: Fix some PTP/IP error messages
camlibs/ptp2: Fuji: camera_trigger_camera: Poll slower
camlibs/ptp2: Don't send final ptp_initiatecapture with Fuji cameras
camlibs/ptp2: Fuji PTP/IP init packet must have length 82
camlibs/ptp2: Fuji: When connecting via PTP/IP try to connect twice
Add add missing ISO values from Canon EOS R6
David Rosca (3):
ptp2: Disconnect PTP/IP connection in camera_exit
filesys: Add gp_filesystem_set_info_dirty
ptp2: Reload object info on PTP_CANON_EOS_CHANGES_TYPE_OBJECTINFO_CHANGE
Gergely Erdélyi (1):
Add model detection for Fujifilm X100V (#627)
Hans Ulrich Niedermann (112):
Fix typos found by codespell in *.m4
gitignore: Anchor patterns at top_srcdir
gitignore: Add /tmpwrk*/
Fix grammar and spelling in comments and text
Remove obsolete gp2ddb (gphoto2 device database) idea
configure.ac: Avoid po-directories error
Use AC_SEARCH_LIBS to find where mmap(2) is defined
gh workflows: remove requirements for flex, bison
INSTALL: add gtk-doc Debian package name
Document using GP_CAMLIB_SET on make command line
Run libgphoto2_port tests only AFTER building iolibs
Leave comment in Makefile.am about subdir sequence
gitignore: Add a few forgotten files for in-source-tree builds
Remove remnants of OS/2 support
use the proper word
fix broken check of CPP macro definition after #elif
Terminate assignment with semicolon instead of comma
Fix parameters being marked unused when they are being used
Use proper format string for FDHANDLE type
comment out unused constant/variable definition
Use consistently named non-reserved CPP macros for header files
Use a single definition of BayerTile enum
utf-8: Convert ISO-8859 text to UTF-8
utf-8: change xml/html from iso-8859-1 to utf-8
utf-8: fix weird minus/hyphen character
utf-8: convert old camlibs/**/*.txt files from latin1
Remove cvs-tags make target
HACKING.md: Update include guard macro name
Use same autoconf requirement in libgphoto2_port
Group iolib list related definitions
Avoid accidental include file name collisions
Rename bayer.[ch] to stv0680-bayer.[ch]
Use common code for defining iolibdir and camlibdir
Make CAMLIBS and IOLIBS fallbacks configurable
Replace --with-foodir= argument by foodir= variable
Fix check output for DEFAULT_CAMLIBS and DEFAULT_IOLIBS
test-camera-list: Add --help output
test-camera-list: Widen column for 4 digit number of camlibs
Remove unused remnants of convenience libltdl
Sync gp-byteorder.m4, supporting 64bit values
Use && and || instead of test arguments -a and -o
Remove rpm packaging
Remove unused code adding x bit to *.sh
Remove linux-hotplug support
Update libphoto2_port/po/de.po to UTF-8 after all
NEWS: converted text from iso to utf-8, removed linux-hotplug
Typo fix: It is _SOURCES, not _SOURCE
example: include <unistd.h> for close(2), unlink(2), getpid(2)
tp6081-dump: include <unistd.h> for close(2), add more static
Here document: Use <<EOF everywhere
sync gp-check-library.m4 between gphoto-m4 dirs
Factor out i18n stuff into separate i18n.h files
Delete old debug comments
library versioning comment without trailing spaces
examples: Force #include beginning with "libgphoto2/"
Rename libgphoto2_port test directory tests
POTFILES.in: Add ptpip, vusb iolibs
POTFILES.in: Add remaining source files using i18n.h
tests: include system headers before gphoto2 headers
tests: declare local variable local
sync gp-gettext-hack.m4 files between gphoto-m4 dirs
HACKING.md: Misc small improvements (grammar, whitespace, etc.)
HACKING.md: Update with C99 being our requirement
HACKING.md: Add note about printf(3) fmt strings using "%m"
gp-camlibs.m4: Fix GP_SLEEP invocation in in Gentoo hack
tests: Remove old pedantic compilation tests
tests: Always define, but only test conditionally
tests: add pedantic compile tests for -std={c2x,c++20}
Make the compiler warnings work with clang, gcc, and others
Update libltdl checks (GP_LIBLTDL)
Update the checks for libjpeg (GP_LIBJPEG)
gitignore for tests/
Use m4_pattern_forbid to find unnoticed GP_ strings
AC_INIT change to arguments on multiple lines
AM_INIT_AUTOMAKE change to indented multi-line form
AC_CONFIG_FILES change to indented multi-line form
m4 quote all macro arguments
gh workflow: install libgd-dev
Consistent checks for asm .symver and version-script
Use the same gettext setup
Use LINGUAS files instead of ALL_LINGUAS variable
remove old libtool/libltdl comments
make gphoto-m4 directories identical
indentation for readability
Remove definition of localedir
Add build tool version comment
Require at least autoconf-2.69
Require at least automake 1.14
Only use a single gphoto-m4/ subdir
Add "mode: autoconf" line to all *.m4 files
with-camlibs: grammar fix for standard set description
with-camlibs: document "everything" set in the --help output
Use non-reserved macro name for generated gphoto-endian.h
Add compiletime-assert.h for COMPILETIME_ASSERT() macro
ptp2/ptp.c: tell emacs the c-file-style
ptp2/ptp.c: Rewrite PTP_CNT_INIT to build with pure C99
remove whitespace around AS_HELP_STRING options
Remove --with-doc-dir and --with-html-dir
rearrange test-pedantic definitions
The pedantic compilation flags might ignore some warnings
RELEASE-HOWTO.md: Change the name of the IRC network
Invoke AC_PROG_SED before using ${SED}
Require at least libtool 2.4.2, use LT_INIT
Remove commented out hack for pre-2001 gettext
Add TODO comment about possibly using gettext's gettext.h
MAINTAINERS: remove rpm builds
Document translating languages outside Translation Project
RELEASE-HOWTO.md: Improve markdown readability
Require gettext >= 0.18.3; remove GP_GETTEXT_HACK
Add pull-translations target
Add test-i18n.c i18n test program
Verbose gettext domain consistency checks
Ingvar Stepanyan (11):
Provide correct search size to memchr
Revamp Sony ISO logic
Fix infinite loop when setting Sony values
Fix overshoot logic for Sony enums
Mark properties without setters as readonly
Add file extension for Sony ARW files
Remove duplicate GP_MIME_ARW
Normalize Sony shutter speeds
Fix memory leak in capture preview for Sony
Add Sony a6600 descriptors
Add few more movie modes for Sony
Jakub Bogusz (1):
updated polish
Jan Kundrát (4):
camlibs/ptp2/ptp-pack.c: Fix aperture on EOS RP
Canon: basic control over Movie Servo AF
Add mode dial positions available on EOS RP or 5Div
Add HDR exposure mode from EOS RP
Jan Tojnar (1):
print-camera-list: clarify help
Jaroslav Škarvada (1):
Fixed build in mingw
Jiri Machalek (1):
mime type update for CR3
Marcus Meissner (242):
2.5.26.1 development starts
define _DARWIN_C_SOURCE for strdup on MacOS. fixes gphoto/libgphoto2#552
also do tar.xz
do not add changecamera mode on the Nikon V1
do not query for PTP_DPC_NIKON_LiveViewStatus
mark nikon v1 as with broken cpature,. also disable the 90c7 event handling as this also crahes the camera. gphoto/libgphoto2#569
on nikon v1, also blank out the nikon specific capture
pass the usb interface to the control pipeline command functions
added 2 new ricoh models from libmtp/libmtp#68
comment out the broken flags again so we get good filesiszes from Samsung Galaxy
dedup a name
revert back to 0 byte PTP strings turning into "" instead of NULL.
detect objects larger than 4gb.
added 64bit file support for Nikon DSLR
add 4gb news
added ricoh theta z1 and v debugging
wait busy also handle the bulb captuzre busyness
actually use the proplist for the size ... gphoto/libgphoto2#582
remove ()
check for -lws2_32 to enable winsock support for ptpip on windows
fix jpeg -> ws232 typo
added fuji finepix s8600
do not error on files larger than 4GB, our ptp getobject method can handle them.
updated
added new changeapplicationmode for Z6 II
added more z6 ii commands
decode new ptp oc nikon codes
added P1000
added liveview zoom area ... left hand value of WidthxHeight
added fuji x-s10 usb ids and debug output
according to http://fujifilm-x.com/global/support/compatibility/software/x-webcam/ this is doing capture preview at least, gphoto/libgphoto2#603
fuji x-s10 in webcam mode
adjust busy wait on fuji preview capture to be 5ms instead of 1ms
added g12 ids
added 2 more nikon event codes
add a workraound to detect card value even if there is no enum.
added EOS M1 Mark II another ID, from timelapse-VIEW
added Sony A7S III
for Sony ALPHA, return a static list of f-numbers if we get "range" for it.
the Nikon Z series movie capture want to wait for the event confirming finished movie capture before proceeding
add event queue check / remove by type functions.
switch to use a remove event by type functiion in nikon capture
reduce wait times in the eos preview loop again.
added powershot g12 debug
decode EOS names for easier debugging.
the getviewfinderdata has 3 arguments... also use 00200000 like in current traces
implemented nikon check event ex
added missing ptp-pack part for nikon_check_event_ex
handle canon ptpip same as nikon ptpip
add eventmode option for canon eos.
hook in nikon geteventex function.
support more MTP ObjectPropDesc types
The EOS M6 Mark II is more similar to the traditional EOS, than the EOS M series.
x-s10 has only webcam mode, no capture mode... fixes gphoto/libgphoto2#603
added fuji x-s10 iopcodes
year is 2021
autopoweroff might be 32bit
refine debug of ptpip responses
skip the unknown data in front of the preview using the 4 byte offset
added sony a7r2 debug
handle deviceinfo changed events in wait_for_event
remove superflous print device info
added streamid parameter to GetStreamInfo
added canon eos m200
blank serial nr
added 9414 getliveview config
added more autoexposuremodes (from eos m6 mark 2)
decode more canon events a bit better
set the cature timeout BEFORE calling captureimage, helps nikon v1
generic initiate capture path:
manage all Z series with te z6 config
report movieprohibit and liveviewprohibit
camlibs/ptp2/ptp-pack.c: Fix get-config iso/aperture for R5. (#621)
seperate prohibits by comma
move some enums to their known places, eos cameranickname is a string
report camera nickname
added OLC 0x12 for EOS M6 Mark2 offsets
adjust more offset for eos m6 mark 2 OLC
added strobofiring, flashcharged, oneshotrawon, autoexposuremodedial
return the correct OK value on success
added liveviewsize configuration for panasonic. set is not workign fully. gphoto/libgphoto2#575
also include x value in config for now
replaced some malloc(x*y) by calloc(x,y), also removed then superflous memset
replaced some malloc(x*y) by calloc(x,y), removed 1 superflous memset
replaced more malloc(x*y) by calloc(x,y)
move around panasonic liveviewdata
use define instead of magic constant
optimize panasonic whitebalance
added more info on panasonic things
implemented getcapturetarget
add more defines for panasonic, use some defines for queries instead of magic values
added movie recording for panasonic lumix
added panasonic deviceprop sniffs
added more sniffed data
added a missing propdesc freeing
added missing propdescfree
added some missing free propdesc and free objectinfo in sony trigger capture
movie is a toggle,. not text
removed a double free (Coverity)
imported from libmtp
removed a superflous argument (Coverity)
updated to current state
fixed a memleak (Coverity)
fixed incorrect NULL check (Coverity)
2.5.27 release
update RELEASE-HOWTO.md
2.5.27.1 dev release
added Konica-Minolta:DiMAGE Z6
move Fujifilm X100V to corect ordered by place
olympus would wait twice the time due to a missing return
added polympus note
also added DSC-RX0M2 here for intiial wait, see gphoto/gphoto2#349
added note about manual focusing not needing delay
added sony a7c
also wait 3 seconds for the A7C
add download when announcing
added #define _DARWIN_C_SOURCE as in other files to avoid implicit
use isoc99 source gphoto/libgphoto2#633
use darwin_c_source ... isoc99 source does not seem to cut it for macos. gphoto/libgphoto2#633
support sony prioritymode, init it to 1.
added d780 aapture settings.
added coolpix b500
added movieexposurecompensation
fix movieexposurebiascompensation
added fuji xpro3
start sigma fp support. gphoto/libgphoto2#537
added sigmafp liveview function, decode sigmafp codes
sigma fp liveview now works
first sucessfull (but not fully correct) capture code for sigma fp
added cleardbsingleimage after capture
implemented getpictfileinfo2ext for sigma fp
implemented capture status, used it to cehck for finished capture
wait a bit longer for image processing
add some more defines, name 9035 camerainfo
pass id to clearimagedb
debug output all data groups for the sigma fp
start the work of allowing sigma fp configuration
first try at shutterspeed for sigma fp
added sonx rx1r m2
added eos m50m2
added sigma fp shutterspeed mapping from SDK table
added sigma fp aperture handling (untested)
added correct property for manualfocus sony
added liveview support to x-t30, but no capture. gphoto/gphoto2#196
added d5000
report that nikons cannot upload
fuji: instead of special casing fuji, just return GP_OK from the fuji branch
added Fuji:Fujifilm GFX 100S
updated NEWS
added eos1300d
allow cancel for bytes_read >= 1MB to cover blob read size in ptp
added a7s data
harmoize the jpeg extractors pointer arithmetic for memchr to be valid C
use the large jump on first try, fall back to single jumps on next loops
added more braces to the new expressions to make clear what we test
make the textual description of syncdatetimeutc also mention UTC
fixed some memory leaks caused by ptp_getobjectinfo imbalanced with ptp_free_objectinfo, see gphoto/libgphoto2#690
temporary allow setting all properties on sony, as ourdetection logic does not seem to be right.
remove memory leaks
report changed config labels during wait
decode value of regular widgets
added property change decoding to EOS and Olympus
more debug, also switch conditions to avoid O(n2) complexity
srtore DevPropCode, needed by current config name matcher
trying to fix eos reader
emit also new values in the config update work (draft work) gphoto/libgphoto2#677
factorize out the jpeg extraction out of data blobs
further factor out the preview saving and meta data setting, as its common
use generic preview.jpg name, not sony for ... fuji
move another preview jpeg extractor to factorized function
factor out regular image additions repeated code
added fuji gfx100 second id reported via SF
for sony preview capture use a timeout of 1 second
replaced R2 -> RP
added sony zv-e10
added sony zv e10
pass lenght of datablobs into the directory compare function to avoid overreads
harden the directory compare function against overreads
handle busy errors on getobjectinfo
turn zoom widget into a range.
check for errors from widget functions
merged music-players.h from libmtp
also register port vusb:
start a sample for libfuzz
addred a missing gp_port_info_new
remove 1 warning
added sample-libfuzz ... semi-uable for libfuzzer
added comment how to build fuzzer, and current problems
fixed uninitialized memory access / potential crash
better comments
added u1-u3 for nikon
decode usermode
added z6 note
report file info for debugging
report only if flag set
also handle j1 like v1 ...blank out getevent and imagerecinsdram
report file info
requzest the objectinfo to be loaded
fixed return code from unprepare capture during EOS exit
emit only matching resetuilock if we did setuilock
added j1 issue
fix build after bayer.h change
added pentax k3 II in ptzp mode
only ovbrwrite the PEF header for the K100D.
emit a FILE CHANGED event if the ObjectInfo changes.
some updates
somre more updates
rev the age of the library.
added note for the 1:0:7 rev
add more news
make update-po
updated german
german translation completed
spelling fixes
more speling fixes
more spellchecking
avoid duplicate mvoie / liveview prohibition code strings for translation
simplified after prohibit simplification
added creative auto mode dial setting
sync from libmtp
handle motorola as MTP... fixes gphoto/gphoto2#463
we need to refetch the ptp object after get_folder_from_handle, as that call might invalidatze the poionter
Factored out more things into add_object_to_fs_and_path helper
factory out 2 more instances for standardized object adding
factor out more instances of 'add handle to filesystem'
updated de
fixed an uninit variable
added c++ wrappers
remove superflous ;
avoid crash on freed memory
added pentax k1
added pentax k1
aded sony sx-hx80
added pentax k1 id
added Sony ILCE-7RM4A gphoto/libgphoto2#749
remove dup k1
adjust hex id to have 4 digits, fixes gphoto/libgphoto2#753
fix camera name, fixes gphoto/libgphoto2#754
sync from libmtp
added more NEWS
keep canon eos remotemode=1 on exit, to keep the display on
2.5.28 release
Ole Dalgaard (1):
Add Nikon Zfc support
Ondrej Zary (3):
sierra: Add workaround for Epson PhotoPC 500 at higher speeds
sierra: Add SIERRA_PACKET_CANCEL (0x18) as valid resonse
sierra: Add curly braces to improve code readability
Pino Toscano (2):
directory: use statvfs instead of statfs
directory: enable storage_info_func on any platform
Rafał Hirsz (1):
Add _DARWIN_C_SOURCE to most files that use string.h (#615)
Stéphane Aulery (2):
updated french
updated french
Sven Killig (1):
added PowerShot exposure compensation values
Wolfram Sang (1):
ptp2: add Canon IXUS 132
Yuri Chornoivan (1):
updated ukrainian
Zeranoe (2):
Remove stdint.h shim (#625)
Fix byte swap calls (#626)
avisecag (1):
Added USB ID for Nikon Z6_2 (#583)
fahrion2600 (2):
Recognize Nikon Z 7II and fix VendorExtensionID over PTP/IP (#605)
Also set params->controlmode in _put_Nikon_ControlMode (#606)
gregmag (1):
Update library.c
tomsch1 (2):
Update library.c
Create sony-a7r3a.txt
trilader (1):
Add support for reading Canon EOS MovieMode switch and setting it on/off
Originally reported in #689 (comment), copying here for a separate tracking issue: most usages of
ptp_getobjectinfoseem to leak the resulting heap-allocated strings and must callptp_freeobjectinfo.Or, perhaps a more systematic fix is necessary - for example,
PTPObjectInfocould be changed to contain fixed-length char arrays and sidestep this problem altogether.Maybe I missed a few, but I haven't seem examples of that.
E.g. from the same file, the Fuji switch below:
libgphoto2/camlibs/ptp2/library.c
Line 3692 in fcdf031
oiwill also be leaked like the Sony one.Same for this one:
libgphoto2/camlibs/ptp2/library.c
Line 4489 in fcdf031
Or here:
libgphoto2/camlibs/ptp2/library.c
Line 4893 in fcdf031
...and there are many others.
The text was updated successfully, but these errors were encountered: