TriBITS: Problem with Mac using BSD stat vs. Gnu stat

[TBird2001]

Bug Report

Description

This is specific to Release 13.0.1. I just ran into the exact same bug that was reported in Issue #7881. Looking at the history, the fix went into the master branch after the 13.0.0 release, and it never got ported to the release-13 branch. I can work around it for now, but it would be nice to have it in a Trilinos release.

[TBird2001]

I forgot to tag @bartlettroscoe in the report. Doing that here.

[bartlettroscoe]

CC: @TBird2001

@trilinos/framework, is putting out another minor release of Trilinos 13.0.z an option? It occurred to me that I should ask this first before updating the Trilinos 13.0 branch to address this.

[bartlettroscoe]

@trilinos/framework, is putting out another minor release of Trilinos 13.0.z an option? It occurred to me that I should ask this first before updating the Trilinos 13.0 branch to address this.

@trilinos/framework, is putting out Trilinos 13.0.2 after this patch an option? This is an important internal customer asking for this.

[jwillenbring]

@trilinos/framework, is putting out Trilinos 13.0.2 after this patch an option? This is an important internal customer asking for this.

Yes. There is actually another potential driver for doing this soon.

[bartlettroscoe]

Yes. There is actually another potential driver for doing this soon.

@jwillenbring, okay, I will put in the patch.

On a related note, is there a plan for a Max OSX PR or develop-to-master build for Trilinos? If that build also did an install as a byproduct of the build (like is done for the ATDM Trilinos builds), then this defect would have been caught as part of regular testing before getting out into a release.

[jwillenbring]

On a related note, is there a plan for a Max OSX PR or develop-to-master build for Trilinos? If that build also did an install as a byproduct of the build (like is done for the ATDM Trilinos builds), then this defect would have been caught as part of regular testing before getting out into a release.

There are some machines now where we could potentially run some tests. Adding a PR build for OSX has not been prioritized highly enough to make it happen. Given the scalability issues with the current system, adding PR builds is carefully considered. I would be interested to learn about any strong internal drivers for such a build.

[bartlettroscoe]

Given the scalability issues with the current system, adding PR builds is carefully considered.

@jwillenbring, any ETA on what that will be fixed so that if just one test fails in one PR build you don't rerun all the PR builds?

I would be interested to learn about any strong internal drivers for such a build.

A lot of Trilinos developers and users do development on their Mac OSX laptops.

[jwillenbring]

@jwillenbring, any ETA on what that will be fixed so that if just one test fails in one PR build you don't rerun all the PR builds?

I don't anticipate that this issue will ever be fixed in the current tool due to the fact that all development activity seems to be focused on the next generation tool. From what I understand, the next generation tool is moving slowly due to other commitments. I am interested in exploring completely different approaches that could help with this issue, but do not currently have staffing to explore those ideas.

[bartlettroscoe]

So the problem that I am having here is that the branch trilinos-release-13-0-branch was based on a version of TriBITS from May 22, 2020 (almost a year ago) from snapshotted TriBITS commit TriBITSPub/TriBITS@52acd38 pulled into Trilinos in Trilinos commit b202385. There have been a lot of changes to TriBITS in that time including some backward incompatible changes:

$ cd TriBITS/

$ git diff github/master --not 52acd38 -- tribits/ReleaseNotes.txt

  Release Notes for TriBITS
 ----------------------------------------
 
+2021/03/12:
+
+(*) MAJOR: Upgrade minimum required CMake version from 3.10 to 3.17.  Existing
+    TriBITS projects that have already upgraded to require CMake 3.17+ should not
+    notice any major changes due to this change.
+
+2020/11/12:
+
+*) MAJOR: The default for `<Project>_ENABLE_EXPLICIT_INSTANTIATION` (ETI) was
+   changed from `OFF` to `ON`.  This was turned on in practice for almost all
+   users of Trilinos so while this change technically breaks backward
+   compatibility of TriBITS, in practice, this will likely not impact any
+   important customers.  (And new customers of Trilinos and related TriBITS
+   projects will enjoy the benefits of ETI by default.  See
+   trilinos/Trilinos#8130.)
+
+2020/10/08:
+
+(*) MAJOR: Tests defined with `TRIBITS_ADD_TEST()` and
+    `TRIBITS_ADD_ADVANCED_TEST()` with `NUM_MPI_PROCS > 1` will now not be
+    added for non-MPI TPL_ENABLE_MPI=OFF configurations.  Before, the test
+    would get added and basically ignore the value of `NUM_MPI_PROCS`.  This
+    change together with the change to move to using `add_test(NAME <name>
+    COMMAND <command>)` mostly restores backward compatibility for projects
+    using TriBITS.  For more details, see trilinos/Trilinos#8110.
+
+2020/09/28:
+
+(*) MAJOR: Tests defined with `TRIBITS_ADD_TEST()` and
+    `TRIBITS_ADD_ADVANCED_TEST()` have been updated from using
+    `add_test(<name> <command>)` to using `add_test(NAME <name> COMMAND
+    <command>)`.  This now causes CMake to error out if there is more than one
+    test with the same name in the same directory.  Before, CMake would allow
+    it (but the behavior in that case was undocumented and undefined).  For
+    more details, see trilinos/Trilinos#8110.
+
+2020/06/16:
+
+(*) MINOR: The variables <Package>_FORTRAN_COMPILER and
+    <Package>_FORTRAN_FLAGS in the <Package>Config.cmake files (in the build
+    dir and the install dir) are deprecated.  Please use the new variables
+    <Package>_Fortran_COMPILER and <Package>_Fortran_FLAGS.  (Justification:
+    In raw CMake, the compiler is called 'Fortran', not 'FORTRAN'.  Also, the
+    name 'Fortran' is already used in the <Project>Config.cmake file.)

Therefore, I can't just copy over a new TriBITS.

I will therefore have to try to surgically pull over just the changes to TriBITS to address this issue. I will have to manually test with Linux but I can't test on a Mac OSX machine so someone will have to do that for me to ensure this is fixed.

[TBird2001]

I will therefore have to try to surgically pull over just the changes to TriBITS to address this issue. I will have to manually test with Linux but I can't test on a Mac OSX machine so someone will have to do that for me to ensure this is fixed.

@bartlettroscoe, I can do the testing on my Mac, if you don't get any other volunteers. Just let me know the best way to get access to the changes.

[bartlettroscoe]

@TBird2001, the fixes should be in the commit in the branch in PR #9110. (That PR must pass PR testing and be approved by at least one Trilinos developer before it can be merged to the Trilinos 13.0 branch.) Just pull that branch from my fork and give it a try. (I can provide exact instructions on how to do that if needed.)

[TBird2001]

@bartlettroscoe, I just tested it on my Mac, and the installation went smoothly. The fix appears to have worked!

[github-actions]

This issue has had no activity for 365 days and is marked for closure. It will be closed after an additional 30 days of inactivity.
If you would like to keep this issue open please add a comment and/or remove the MARKED_FOR_CLOSURE label.
If this issue should be kept open even with no activity beyond the time limits you can add the label DO_NOT_AUTOCLOSE.
If it is ok for this issue to be closed, feel free to go ahead and close it. Please do not add any comments or change any labels or otherwise touch this issue unless your intention is to reset the inactivity counter for an additional year.

[github-actions]

This issue was closed due to inactivity for 395 days.