Fix bad configure tests #42756
Fix bad configure tests #42756
Conversation
There was an extra -c in the CMAKE_REQUIRED_FLAGS set for testing HAVE_UNW_GET_ACCESSORS and HAVE_UNW_GET_SAVE_LOC that was breaking build of coreclr under homebrew. The option was somehow making these checks behave on ARM Linux, eveb though it is not clear to me why, as it was just causing this option to be passed to the compiler twice at different positions of the command line of the cmake tests. This change fixes it by using check_symbol_exists instead of check_c_source_compiles, since just removing the duplicite -c was resulting in the check failing on ARM / ARM64 Linux due to a missing symbol from libunwind during linking.
|
cc: @omajid |
|
Awesome! What's the ETA of this becoming available in |
See: dotnet#1744 This is a backport of the fix that was merged into runtime for master: dotnet/runtime#42756
|
Will this be backported for 5.0.0? |
If dotnet/source-build#1789 is merged, the next monthly release (mid-October? mid-November?) tag will include the fix. |
It's too late for the October tag unless there's a build reset (not expected), but if you're ok using a commit rather than a tag (I think you mentioned this is ok at some point?), you can start using it immediately after dotnet/source-build#1789 is merged. |
|
Cool, I'll definitely update the Homebrew Formula pointing to the merge commit SHA once dotnet/source-build#1789 is merged. |
|
/backport to release/5.0 |
|
Started backporting to release/5.0: https://github.com/dotnet/runtime/actions/runs/289510941 |
|
I figure whether or not it'll make 5.0.0 itself, might as well kick it off. Saw people using that incantation in other PRs. 😛 |
See: #1744 This is a backport of the fix that was merged into runtime for master: dotnet/runtime#42756
ghost
locked as resolved and limited conversation to collaborators
There was an extra -c in the CMAKE_REQUIRED_FLAGS set for testing
HAVE_UNW_GET_ACCESSORS and HAVE_UNW_GET_SAVE_LOC that was breaking
build of coreclr under homebrew. The option was somehow making
these checks behave on ARM Linux, eveb though it is not clear to
me why, as it was just causing this option to be passed to the
compiler twice at different positions of the command line of
the cmake tests.
This change fixes it by using check_symbol_exists instead of
check_c_source_compiles, since just removing the duplicite -c
was resulting in the check failing on ARM / ARM64 Linux due
to a missing symbol from libunwind during linking.