[PATCH] D81408: [builtins] Improve compatibility with 16 bit targets

[Anatoly Trosinenko via Phabricator via cfe-commits]

atrosinenko added a subscriber: goncharov.
atrosinenko added a comment.

In D81408#2104572 <https://reviews.llvm.org/D81408#2104572>, @efriedma wrote:

> > Those flaky test failures seems to be due to ld.lld being not built from source as part of testing compiler-rt/-only patches.
>
> That should be something we can fix in the build system.  compiler-rt/test/CMakeLists.txt has a list of executables which the tests depend on. If that list isn't complete, it should be updated.


@efriedma

Agree, but another question: is it acceptable to have them disabled when retesting compiler-rt? So, maybe lld subproject have to be explicitly enabled as well for `compiler-rt`-only tests...

@goncharov

Meanwhile, looking at the latest Buildkite build <https://buildkite.com/llvm-project/premerge-checks/builds/898> for this review, we have:

name=artifacts/CMakeCache.txt (Linux)
  ...
  LLVM_ENABLE_PROJECTS:STRING=clang;compiler-rt;llvm
  ...

and

name=artifacts/CMakeCache.txt (Windows)
  ...
  LLVM_ENABLE_PROJECTS:STRING=clang;llvm
  ...

So, isn't Windows build here just wasting CPU time?

Another my concern here is that the build plan pre-merge checks (buildkite) <https://reviews.llvm.org/harbormaster/plan/5/> is expressed with just a single "Make HTTP request" step with "When Complete: Wait For Message" (unlike the pre-merge checks <https://reviews.llvm.org/harbormaster/plan/4/> used for the initial upload). So, this build <https://reviews.llvm.org/harbormaster/build/73646/> failed like this (see the message tab):

  216557	merge_guards_bot	fail	Consumed	Thu, Jun 18, 16:13
  216556	merge_guards_bot	working		Thu, Jun 18, 16:13
  216538	merge_guards_bot	working		Thu, Jun 18, 15:51

In D81282: [builtins] Move more float128-related helpers to GENERIC_TF_SOURCES list <https://reviews.llvm.org/D81282>, on the other hand, there is seemingly passed build <https://reviews.llvm.org/harbormaster/build/73581/>. Its messages are

  216370	merge_guards_bot	fail		Thu, Jun 18, 13:28
  216365	merge_guards_bot	working		Thu, Jun 18, 13:25
  216343	merge_guards_bot	pass	Consumed	Thu, Jun 18, 12:53
  216342	merge_guards_bot	working		Thu, Jun 18, 12:52

So, it looks like now the `compiler-rt`-only pre-merge check cannot pass at all, but can be **marked** Passed **on re-upload only** based on what builder (Linux or Windows) will send its reply first.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D81408/new/

https://reviews.llvm.org/D81408