[libcxx] Do not redeclare lgamma_r when targeting the LLVM C library
#102036
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jhuber6
requested review from
ldionne,
mordante and
philnik777
and removed request for
a team
llvmbot
added
the
libc++
|
@llvm/pr-subscribers-libcxx Author: Joseph Huber (jhuber6) ChangesSummary: Full diff: https://github.com/llvm/llvm-project/pull/102036.diff 2 Files Affected:
diff --git a/libcxx/include/__configuration/platform.h b/libcxx/include/__configuration/platform.h
index 27f68d04e8a8d..a53ec7d5f7b4a 100644
--- a/libcxx/include/__configuration/platform.h
+++ b/libcxx/include/__configuration/platform.h
@@ -38,7 +38,9 @@
# else
# define _LIBCPP_GLIBC_PREREQ(a, b) 0
# endif // defined(__GLIBC_PREREQ)
-#endif // defined(__linux__)
+#elif defined(__AMDGPU__) || defined(__NVPTX__)
+# include <features.h>
+#endif
#ifndef __BYTE_ORDER__
# error \
diff --git a/libcxx/include/__random/binomial_distribution.h b/libcxx/include/__random/binomial_distribution.h
index e8774bb8d67ee..3f19746bae238 100644
--- a/libcxx/include/__random/binomial_distribution.h
+++ b/libcxx/include/__random/binomial_distribution.h
@@ -97,7 +97,8 @@ class _LIBCPP_TEMPLATE_VIS binomial_distribution {
}
};
-#ifndef _LIBCPP_MSVCRT_LIKE
+// The LLVM C library provides this with conflicting `noexcept` attributes.
+#if !defined(_LIBCPP_MSVCRT_LIKE) && !defined(__LLVM_LIBC__)
extern "C" double lgamma_r(double, int*);
#endif
|
ldionne
requested changes
Aug 14, 2024
| @@ -38,7 +38,9 @@ | |||
| # else | |||
| # define _LIBCPP_GLIBC_PREREQ(a, b) 0 | |||
| # endif // defined(__GLIBC_PREREQ) | |||
| #endif // defined(__linux__) | |||
| #elif defined(__AMDGPU__) || defined(__NVPTX__) | |||
There was a problem hiding this comment.
This makes me wonder if we should instead do this:
// <features.h> defines macros that allows detecting the libc we're using
#if __has_include(<features.h>)
# include <features.h>
#endif
#if defined(__GLIBC_PREREQ)
// etc...
#endif
Can you try that out and see if that breaks anything in the CI?
Summary: We use `lgamma_r` for the random normal distrubtion support. In this code we redeclare it, which caused issues with the LLVM C library as this function has the `noexcept` flag. Also make sure we include `features.h` when using the GPU so it gets this as well.
|
ping |
1 similar comment
|
ping |
ldionne
approved these changes
Aug 27, 2024
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/72/builds/2687 Here is the relevant piece of the build log for the reference |
ldionne
reviewed
Jan 15, 2025
Comment on lines
+33
to
+34
| // To detect which libc we're using | ||
| #if __has_include(<features.h>) |
There was a problem hiding this comment.
When pre-qualifying LLVM 20, we just found out that this breaks some code. Indeed, folks are free to have a header named <features.h> available on their include paths, and in fact it seems not uncommon to have that. So the __has_include(<features.h>) is going to lie, and the #include <features.h> is going to pick up the wrong one anyway.
We'd need to find another header we can include to provide this version macro. Since C lacks a standardized header to provide version stuff (what a shame), we often rely on using some minimal header like <iso646.h> to get implementation-specific values. That should work for glibc, and llvm-libc could be changed to provide a <iso646.h> header. Also, all of llvm-libc's headers should define the feature macros like __LLVM_LIBC__, which I think is not the case right now.
ldionne
added a commit
to ldionne/llvm-project
that referenced
this pull request
Feb 3, 2025
…library (llvm#102036)" This reverts commit 5f2389d. That commit started checking whether <features.h> was a valid include, however codebases are free to have such a header on their search path, which breaks compilation. LLVM libc should instead provide a more standard way of getting configuration macros like __LLVM_LIBC__.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
We use
lgamma_rfor the random normal distribution support. In thiscode we redeclare it, which caused issues with the LLVM C library as
this function is marked
noexceptin LLVM libc. Also make sure we includefeatures.hwhen using the GPU so it gets this as well.