Adjust cmake to handle changes to Max SDK from 8.2 onwards #32
Conversation
|
Should I be building against the 8.2 SDK or the 8.0.3 SDK? |
|
My hope is that it will work with both 🤞 |
|
Okay I shall test here and see what happens. 🙏 |
|
Builds are failing for me because the old HISSTools are pinned. I've suggested a PR to fix this in flucoma-core here. |
|
It's probably easier for dev purposes to use a local copy of HISSTools and set that like the other dependencies via a CMake cache variable (it's called |
|
Unless I am missing something here, I still think that it is necessary to push HISSTools forward in order to update the Max build system to make fat binaries. |
|
You are missing something, insofar as the GIT_TAG for HISSTools in the flucoma-core CMake is only relevant when relying on an automatic download. For the purposes of testing this, using a local HISSTools means we can move the branch around as we like without changing the CMake (i.e., all I had to do was go to my local HISSTools and checkout main). So, yes, for deployment, we'll need to change the CMake but for making sure this change works both for you on the new SDK w/ ARM and for older SDKs on x64, it's not necessary. |
|
Okay I'll test this tomorrow to see how it builds :) Should I just use an invocation I know works (passing FLUID_ARCH etc) as a test case? |
|
Given the developments over @ flucoma/flucoma-core#28 it seems like this is behaving now? |
|
Great. If it works for @tremblap too, then it looks like we're in good shape |
|
It seems to work on sdk7 but suddenly I have 167 targets instead of the usual 114. Is it normal? |
|
Seems like a strange increase. I think you can list targets with ninja by and see if there's anytying confusing in there |
|
with the new version I get: |
|
with the old version - deleting build and ninja cleaning I get 114 targets, the following:
|
|
similar list, but as if the ninja command triggered a lot more compiles... is there a quick way to time the compilation process? it felt longer but what do I know. I'll explore the folder to see if anything gets naughtily put somewhere. |
|
btw my process is to merge this on tip of origin/main |
I'll take your word for it 😄
This is probably due to the new need to buld commonsyms.c alongside the objects (which we've always had to do on Windows). Pre SDK 8.2 it was built into the
Does |
nope. it is longer. i can live with it if objects are working as before. Cursory testing seems to imply it is, so I'd be happy to merge in main and continue to plough. Saavy? |
|
I have almost always had 167 targets and sometimes 114. It doesn't seem to affect the success of the build. |
|
Shall I merge then? |
|
suspense... |
Merge! 💯 🚀 🚀 🚀 🚀 🚀 🚀 |
* Adjust cmake to handle changes to Max SDK from 8.2 onwards * correct typo
* Adjust cmake to handle changes to Max SDK from 8.2 onwards * correct typo
* Adjust cmake to handle changes to Max SDK from 8.2 onwards * correct typo
* incorporate Ted's changes and recommendations Revert "fix pre/post cmake" This reverts commit 7d04ffb. Create does-it-build.yml does it build? consume SDK curl the 8.0.3 SDK add a ls stage for sanity remove cmake flags for OSX architectures move list cmd add working directories for steps that urn in build use macos as runner use ninja for build process add specific name for build will write a text file named foo to the home directory see if we can headless max patchers also build on the ci branch Buf2List: tidy up and add a 'buffer' message for immediate processing Revert "fat binaries" This reverts commit 35ef645. Revert "build script" This reverts commit 1ee3f5a. # Conflicts: # create.sh Initialize dumpDictionary pointer in MaxWrapper explicitly, as this is assumed in destructor removed dangling script naming to buf2list corrected buf2list use in mlpclass corrected interface of mds to make it more interactive updated buf2list2buf helpfiles with doc dial back automatic patch testing in workflow * Adjust cmake to handle changes to Max SDK from 8.2 onwards (#32) * Adjust cmake to handle changes to Max SDK from 8.2 onwards * correct typo * Update macos-build-test.yml use the latest sdk from github * you actually have to git clone... * only build ampslice for now * test releasing * make names + commits * give a proper asset name * use ref for tag * Update macos-build-test.yml * spell prerelease properly * change convention for naming n stuff * Update macos-build-test.yml * add windows-build script * quotes around path for windows * test a composite workflow * update composite * make a nightly action * now check to see if we can build and see an external * update infrastructure * check with cached outputs too * pre check in the right place * dont check a non-existant file * remove nightly cache workflow * build and combine * remove bad line * do some sanity checks for cleaning * inspect what is going on * cleanup ls -R and change to latest macos * see if we can copy the entire externals folder * check * now build fat and see * remove mistake space * get uname * get clang version * get system info * try some silicon black magic * use custom core * use https * check if core is there * core is only one folder up * make a release again * make externals early * 🩴 :burger: * pass DCMAKE_APPLE flag * full install of max * make release builds on windows * remove pdbs from release-packaging * dont remove pdb files * fix paths because they have spaces 👿 * try and deal with the hell that is windows * now release it! * try again * only build on dev and ci branches * now with DDOCS=ON * ddocs off * provide a more detailed tag for nightly builds * also make docs * use blas compilation branch (temporarily) * fix malformed paths * missing -D * take max_ref and merge with final compile * delete nightly before remaking it * delete extraneous experimental patch Co-authored-by: James Bradbury <jamesbradbury93@gmail.com> Co-authored-by: weefuzzy <gungwho@gmail.com>
* incorporate Ted's changes and recommendations Revert "fix pre/post cmake" This reverts commit 7d04ffb. Create does-it-build.yml does it build? consume SDK curl the 8.0.3 SDK add a ls stage for sanity remove cmake flags for OSX architectures move list cmd add working directories for steps that urn in build use macos as runner use ninja for build process add specific name for build will write a text file named foo to the home directory see if we can headless max patchers also build on the ci branch Buf2List: tidy up and add a 'buffer' message for immediate processing Revert "fat binaries" This reverts commit 35ef645. Revert "build script" This reverts commit 1ee3f5a. # Conflicts: # create.sh Initialize dumpDictionary pointer in MaxWrapper explicitly, as this is assumed in destructor removed dangling script naming to buf2list corrected buf2list use in mlpclass corrected interface of mds to make it more interactive updated buf2list2buf helpfiles with doc dial back automatic patch testing in workflow * Adjust cmake to handle changes to Max SDK from 8.2 onwards (#32) * Adjust cmake to handle changes to Max SDK from 8.2 onwards * correct typo * Update macos-build-test.yml use the latest sdk from github * you actually have to git clone... * only build ampslice for now * test releasing * make names + commits * give a proper asset name * use ref for tag * Update macos-build-test.yml * spell prerelease properly * change convention for naming n stuff * Update macos-build-test.yml * add windows-build script * quotes around path for windows * test a composite workflow * update composite * make a nightly action * now check to see if we can build and see an external * update infrastructure * check with cached outputs too * pre check in the right place * dont check a non-existant file * remove nightly cache workflow * build and combine * remove bad line * do some sanity checks for cleaning * inspect what is going on * cleanup ls -R and change to latest macos * see if we can copy the entire externals folder * check * now build fat and see * remove mistake space * get uname * get clang version * get system info * try some silicon black magic * use custom core * use https * check if core is there * core is only one folder up * make a release again * make externals early * 🩴 :burger: * pass DCMAKE_APPLE flag * full install of max * make release builds on windows * remove pdbs from release-packaging * dont remove pdb files * fix paths because they have spaces 👿 * try and deal with the hell that is windows * now release it! * try again * only build on dev and ci branches * now with DDOCS=ON * ddocs off * provide a more detailed tag for nightly builds * also make docs * use blas compilation branch (temporarily) * fix malformed paths * missing -D * take max_ref and merge with final compile * delete nightly before remaking it * delete extraneous experimental patch Co-authored-by: James Bradbury <jamesbradbury93@gmail.com> Co-authored-by: weefuzzy <gungwho@gmail.com>
* copy the misc folder in installation target * Add relative path file writing for data objects In principle consistent with other Max objects (which aren't wholly consistent in edge cases) * reset points if dict is successfully parsed * crediting (#78) crediting the new team * update kdtree help file * add fluid.stats~ and change category to statistical analysis * move stft and add fluid.grid~ * grid out help file real estate * adds fluid.plotter to the autocomplete objectlist * adds fluid.waveform~ to objectlist * first tab beautified * reorder latency message * getlatency after changing preset * Transients Help File (#82) * grid out help file * update transients~ help file * ci/automatic building (#29) * incorporate Ted's changes and recommendations Revert "fix pre/post cmake" This reverts commit 7d04ffb. Create does-it-build.yml does it build? consume SDK curl the 8.0.3 SDK add a ls stage for sanity remove cmake flags for OSX architectures move list cmd add working directories for steps that urn in build use macos as runner use ninja for build process add specific name for build will write a text file named foo to the home directory see if we can headless max patchers also build on the ci branch Buf2List: tidy up and add a 'buffer' message for immediate processing Revert "fat binaries" This reverts commit 35ef645. Revert "build script" This reverts commit 1ee3f5a. # Conflicts: # create.sh Initialize dumpDictionary pointer in MaxWrapper explicitly, as this is assumed in destructor removed dangling script naming to buf2list corrected buf2list use in mlpclass corrected interface of mds to make it more interactive updated buf2list2buf helpfiles with doc dial back automatic patch testing in workflow * Adjust cmake to handle changes to Max SDK from 8.2 onwards (#32) * Adjust cmake to handle changes to Max SDK from 8.2 onwards * correct typo * Update macos-build-test.yml use the latest sdk from github * you actually have to git clone... * only build ampslice for now * test releasing * make names + commits * give a proper asset name * use ref for tag * Update macos-build-test.yml * spell prerelease properly * change convention for naming n stuff * Update macos-build-test.yml * add windows-build script * quotes around path for windows * test a composite workflow * update composite * make a nightly action * now check to see if we can build and see an external * update infrastructure * check with cached outputs too * pre check in the right place * dont check a non-existant file * remove nightly cache workflow * build and combine * remove bad line * do some sanity checks for cleaning * inspect what is going on * cleanup ls -R and change to latest macos * see if we can copy the entire externals folder * check * now build fat and see * remove mistake space * get uname * get clang version * get system info * try some silicon black magic * use custom core * use https * check if core is there * core is only one folder up * make a release again * make externals early * 🩴 :burger: * pass DCMAKE_APPLE flag * full install of max * make release builds on windows * remove pdbs from release-packaging * dont remove pdb files * fix paths because they have spaces 👿 * try and deal with the hell that is windows * now release it! * try again * only build on dev and ci branches * now with DDOCS=ON * ddocs off * provide a more detailed tag for nightly builds * also make docs * use blas compilation branch (temporarily) * fix malformed paths * missing -D * take max_ref and merge with final compile * delete nightly before remaking it * delete extraneous experimental patch Co-authored-by: James Bradbury <jamesbradbury93@gmail.com> Co-authored-by: weefuzzy <gungwho@gmail.com> * dont use BLAS branch (fixed in pr#67) * remove dangling line * use dev branch of flucoma core for ci * cleanup english and patching on tab 2 * deferlow on file load (was failing before) * update cosmetics * point to /reference/sines in the learn platform not ampslice * add learn and flucoma.org links * apply 10x10 grid * cancel previous runs if a new push is made * Only need 1 outlet * update outlet assistance * update outlet 1 assist * change tag name of nightly to be versionless * Remove per-PR merge checks * update bufselect * add links to top of file * initial cleanup * patcher for linking to learn articles more easily * changes to chroma help file * remove spaces from installation path name (#90) * fix references to spaced package name in CI * sorted the package-info with a few fun anachronicisms * missed a space in macbuild target * only do a release on dev branch * only do a release on dev branch * trigger build on PR * remove per PR building * remove release checking * add a deferlow to fluid.learn * provide a small musical example on first tab * add chroma frequency calculation * replace list.change with zl.change * human unit test maxpat * make a soundful demonstration on each tab * fix incorrect refrences to previous help files * update positioning of online reference/flucoma link * Fix fluid.learn abstraction not displaying properly (#101) * initial cleanup * changes to chroma help file * add a deferlow to fluid.learn * provide a small musical example on first tab * add chroma frequency calculation * replace list.change with zl.change * make a soundful demonstration on each tab * fix incorrect refrences to previous help files * update positioning of online reference/flucoma link * update fluid.learn patch to render more reliably * Load media files from package abstraction (#102) * a preliminary go at the bufloader * tighten regex and implement auto loading * add second tab for curating statistics with bufselect * add cosmetic upgrades to second tab and neaten first tab * fix loading woes with fluid.learn * more cosmetic upgrades to patch * cosmetic upgrades to chroma * [Review Changes] BufChroma Help File (#98) * update fluid.bufchroma~ help file * Fix fluid.learn abstraction not displaying properly (#101) * initial cleanup * changes to chroma help file * add a deferlow to fluid.learn * provide a small musical example on first tab * add chroma frequency calculation * replace list.change with zl.change * make a soundful demonstration on each tab * fix incorrect refrences to previous help files * update positioning of online reference/flucoma link * update fluid.learn patch to render more reliably * Load media files from package abstraction (#102) * a preliminary go at the bufloader * tighten regex and implement auto loading * stripping back of first tab * an interactive playback mechanism with chroma data * chroma lookup * add more detail tab * polish patch * layout fixes for help patch * [Review Changes] KDTree Help File (#92) * more cleanup on fluid.kdtree * separate tabs * add analysis data * cleanup tab 1 * add presliced corpus points * add a tab simplifying the explanation * cosmetic upgrades * add numbered instructions * [CI] Refactor and use composite actions (#106) * use flucoma actions * use composite actions to build windows * remove duplicate env setu * use maxdocs composite action * do CI on every ci located branch * Re-enable building docs as part of ALL * [CI] Fat Binaries (#107) * do CI on every ci located branch * bump to v3 * [CI] Actions@v4 (#108) * do CI on every ci located branch * use v4 of actions * update docs * Re-enable building docs as part of ALL * [CI] Fat Binaries (#107) * do CI on every ci located branch * bump to v3 * [CI] Actions@v4 (#108) * do CI on every ci located branch * use v4 of actions * update docs * indentation * remove duplicate * typo * [CI] Fix Documentation Build (#115) * fix ci docs component * make release all the time * use ubuntu latest for building docs * [Enhance] Update fluid.waveform~ (#116) * update to translational api layer * use new api in help file * buf -> buffer * alias indicesbuffer to addmarkers * update help to reflect internal api * singular for methods * update waveform help * fix pluarlisation of methods and arguments * update help file for fluid.waveform * various cosmetic upgrades (#103) * BufNNDSVD Help File (#88) * update bufnndsvd~ helpfile * add playback to the help file * [Docs] Fix Layout issues in BufChroma help (#117) * fixes layout issues of bufchroma * fix bufcompose load issue * Update fluid.sines~.maxhelp * update dataset help file (#120) * labelset cosmetics * remove frombuffer tab * [Docs] Blocking Help Tab (#121) * add back blocking attribute * add attribute blocking to bufnndsvd * add tab to bufstft * add tab to bufselect * change release action * [Docs] BufLoudness Help (#122) * griddify help file * remove js starter * experiment with waveform-based interface * first tab done (possibly) * finish multichannel tab * PAs changes * [Docs] Loudness Help File (#118) * shell out loudness second tab * cosmetic updates to second tab * update second tab * PA suggestions * typo * more typos * [Docs] Audiotransport Help File (#105) * work on audiotransport help file * add fluid.learn abstraction * cosmetic updates * [Enhance] Retrieve toy DataSets (#126) * pull datasets from the location specificed in PR 88 of core * pull datasets to misc * also add an abstraction for loading datasets * update help file with new asset names * add misc to ignore * copy from Data instead of DataSets * copy resources on cmake configure * moved patchers from misc to patchers * delete the files tracked by mistake * update bufchroma to have an adaptive number of chroma (#129) * AmpSlice Help File (#81) * griddify help file * in progress work for explaining thresholds and curves * update bpatchers for providing linkage to sites * update ampslice to be more friendly and clear * add margin to link helper bpatchers (so they display correctly on load) * update help file to accomodate bpatcher changes * make text box way longer than it needs to be. * update cosmetics * update with review advice * update learn subpatchers * shuffle cosmetics * update bufaudiotransport names (#131) * Update nightly.yaml Add workflow dispatch for manual launch * correct docs path variable in CMake * [Docs] Grid Help File (#110) * add datasets that can be reused for examples * update to fluid.grid helpfile * try and make learn load in the right order... * update aesthetics and separation of tabs * cosmetic upgrades * Revert "add datasets that can be reused for examples" This reverts commit 20f58a3. * update dataloader abstraction * fix layout with new dataset loader * Max buffer references weren't being freed safely (ASan) (#132) * update quickstart * [Fix] fluid.plotter~ clear label data (#135) * fix single quotes * remove old var * clear all of the data that we need to clear * re-order precedence of colouring (#136) Co-authored-by: weefuzzy <gungwho@gmail.com> Co-authored-by: tremblap <info@pierrealexandretremblay.com> Co-authored-by: Timo Hoogland <timohoogland@gmail.com> Co-authored-by: James Bradbury <jamesbradbury93@gmail.com> Co-authored-by: Ted Moore <ted@tedmooremusic.com>
commonsyms is no longer included in the
.framework(which was apparently an error), so needs to be built with the mxo on Mac as well as windows nowThis seems to work fine against the v7 SDK. Weirdly it wouldn't work until the flag for including the precompiled headers was moved from the CXXFLAGs into the target compiler options (which is better anyway). Unfortunately, the error you get when this is broken is not at all helpful (it dies trying to find an old carbon header, Files.h)
The directory structure has changed such that what used to be
sources/c74supportcan now besources\/max-sdk-base/c74support.This gives three possible scenarios:
max-sdk-basepathc74supportfolderI'm doing this with an
iftree at the moment, which sucks. It would be nicer to divine the SDK version fromext.hdirectly, but that's a chicken and egg problem (because we have to find it first) .@jamesb93 @tremblap can you see if your respective builds both work against this?