Update Windows CI to not install ninja #4230
Conversation
jhendersonHDF
added
Merge - To 1.14
jhendersonHDF
requested review from
lrknox,
derobins,
byrnHDF,
fortnern,
qkoziol,
vchoi-hdfgroup,
bmribler,
glennsong09,
mattjala and
brtnfld
as code owners
| @@ -135,10 +135,6 @@ jobs: | |||
|
|
|||
| # CMake gets libaec from fetchcontent | |||
|
|
|||
| - name: Install Dependencies (Windows) | |||
There was a problem hiding this comment.
Maybe - Ninja will be needed if we want to use clang (like OneAPI, it is needed for alternative compilers on windows). Note that ninja is required for the binary tests.
There was a problem hiding this comment.
I think we need to be cautious about what checks are required for PRs if they're going to go out and grab things from the internet outside of what's the default on runners. Testing alternative compilers on Windows is great, but probably doesn't need to be a required check.
There was a problem hiding this comment.
Clang and OneAPI on windows are just as important as those compilers for linusx, right?
Unfortunately GH actions for testing are problematic at any time and chocaoately isn't necessarily worse then anything else being used.
The more important metric is time for PRs S.B. less than 25 mins for a workflow.
There was a problem hiding this comment.
Maybe main-cmake.yml would be better named main-cmake-vs.yml or vs-cmake.yml?
There was a problem hiding this comment.
At the very least we should be able to keep a cached installation of ninja and other required things around when we've gotten them once. It also doesn't seem like Ninja is used anywhere for the binary tests. It seems like just the OneAPI Windows action is using it.
There was a problem hiding this comment.
We may also look at https://cmake.org/cmake/help/latest/variable/CMAKE_GENERATOR_TOOLSET.html. Seems like we can use clang and possibly OneAPI as well with VS generators so that we don't need ninja in the first place.
There was a problem hiding this comment.
bintest definitely uses ninja as the generator.
* Take user block into account when returning chunk addresses (#4236) Both H5Dchunk_iter() and H5Dget_chunk_info(_by_coord)() did not take the size of the user block into account when reporting addresses. Since the #1 use of these functions is to root around in the file for the raw data, this is kind of a problem. Fixes GitHub issue #3003 * Fix a minor warning in h5test.c (#4242) * Turn on -Werror for Java in GitHub -Werror workflows (#4243) * Update Windows CI to not install ninja (#4230) * Rework Fortran macros to use the proper code. (#4240) * Correct reference copy for 16 API (#4244) * Determine MPI LOGICAL during build, used in tests. (#4246) * Skip userblock test in chunk_info.c for multi-file VFDs (#4249) * Match generators with real cmake -G output on Windows (#4252) * Add Julia GitHub Actions. (#4123) * Re-revert to using autoreconf in autogen.sh (#4253) We previously tried removing the per-tool invocation of the Autotools and instead simply invoked autoreconf (PR #1906). This was reverted when it turned out that the NAG Fortran compiler had trouble with an undecorated -shared linker flag. It turns out that this is due to a bug in libtool 2.4.2 and earlier. Since this version of libtool is over a decade old, we're un-reverting the change. We've added a release note for anyone who has to build from source on elderly platforms. Fixes #1343 * Rewrite H5T__path_find_real for clarity (#4225) * Move conversion path free logic to helper function * Add tgz extensions on names (#4255) * Remove an error check regarding large cache objects (#4254) * Remove an error check regarding large cache objects In PR#4231 an assert() call was converted to a normal HDF5 error check. It turns out that the original assert() was added by a developer as a way of being alerted that large cache objects existed instead of as a guard against incorrect behavior, making it unnecessary in either debug or release builds. The error check has been removed. * Update RELEASE.txt * File format security issues (#4234) * Add job timeout to cygwin workflow (#4260) * Replace user-define with user-defined (#4261) * Improve the CMake clang -fsanitize=memory flags (#4267) -fsanitize=memory is almost useless without using -fsanitize-memory-track-origins=2 and we shoud probably add -fno-optimize-sibling-calls as well. * Add documentation (H5M) (#4259) * Add documentation (H5P) (#4262) * MPI type correction (#4268) * corrected type for MPI_*_f2c APIs * fixed return type of callback * reset compilation flags of logical test program * Clean up test/cmpd_dtransform.c (#4270) * Clean up test/cmpd_dtransform.c * Fix uninitialized memory warning from sanitizers * FAIL_STACK_ERROR --> TEST_ERROR * Emit output * Delete test file when done * Fix typo * H5Fdelete() --> remove() * Fix uninitialized memory issues in packet table (#4271) * replace deprecated CMAKE_COMPILER_IS_GNU** (#4272) * Prevent stack overflows in H5E__push_stack (#4264) * Minor fixes after merge of file format security fixes (#4263) * Update H5_IS_BUFFER_OVERFLOW to account for 'size' of 0 * Invert a few checks to avoid function call * CHECK --> CHECK_PTR in tmisc.c (#4274) * Add release note for CVE-2017-17507 (#4275) * Update Cygwin installation guide (#4265) * Addresses configuration fortran testing flags (#4276) * turn warnings to errors in fortran configure test * Intel fortran test fix * Merge julia workflows into standard ci format (#4273) * Fix range check in H5_addr_overlap (#4278) When the H5_addr_overlap macro was updated to use H5_RANGE_OVERLAP, it failed to take into account that H5_RANGE_OVERLAP expects the range to be inclusive. This lead to an assertion failure in H5MM_memcpy due to a memcpy operation on overlapping memory. This has been fixed by subtracting 1 from the calculated high bound values passed to H5_RANGE_OVERLAP * Fix potential buffer read overflows in H5PB_read (#4279) H5PB_read previously did not account for the fact that the size of the read it's performing could overflow the page buffer pointer, depending on the calculated offset for the read. This has been fixed by adjusting the size of the read if it's determined that it would overflow the page.
We aren't using ninja in the workflow and it's causing CI actions to fail since https://community.chocolatey.org/api/v2/ is currently down. The OneAPI actions do use Ninja however, so those will likely still fail for now.