Please cut a new release to simplify packaging about mold

[rui314]

@Logarithmus Are you ready for 0.9.2?

[rui314]

Sure. I need to build Gentoo packages to make sure that there's no regression. Please allow me a day or two.

[rui314]

I created a new branch for 0.9.2 and cherrypick only changes that matters to you. Could you confirm that the following branch contains all changes that you need? If it looks OK, I'll tag it as v0.9.2.

https://github.com/rui314/mold/commits/release/0.9.2

[Logarithmus]

@rui314 yesterday I've discovered that multiple tests fail (about 10). So today I'd like to take a record of specific tests which fail and 'll create a PR trying to fix those tests. The package can't be accepted with failing tests.
So please don't create a tag yet.

[rui314]

Got it. For me all tests pass and so do on GitHub Actions, so I wonder what these tests are.

[Logarithmus]

An issue to track failing tests: #67

[Logarithmus]

I created a new branch for 0.9.2 and cherrypick only changes that matters to you. Could you confirm that the following branch contains all changes that you need? If it looks OK, I'll tag it as v0.9.2.

https://github.com/rui314/mold/commits/release/0.9.2

Please, include test fixes for Void into mold/0.9.2 and tag it.

[Logarithmus]

@rui314 UPD 3 tests still fail on my machine, namely:

• copyrel-protected
• ifunc-dso
• run
  I'd like to investigate them before you make a release.

[rui314]

I updated https://github.com/rui314/mold/tree/release/0.9.2 branch.

Could you run each test command with bash -x and share its output?

[Logarithmus]

@rui314

~/Programming/C++/mold fix-tests-again !5 ?1 ❯ bash -x fails/copyrel-protected.sh 
+ set -e
++ dirname fails/copyrel-protected.sh
+ cd fails
++ basename -s .sh fails/copyrel-protected.sh
+ echo -n 'Testing copyrel-protected ... '
Testing copyrel-protected ... ++ pwd
++ basename -s .sh fails/copyrel-protected.sh
+ t=/home/artur/Programming/C++/mold/fails/tmp/copyrel-protected
+ mkdir -p /home/artur/Programming/C++/mold/fails/tmp/copyrel-protected
+ cat
+ cc -o /home/artur/Programming/C++/mold/fails/tmp/copyrel-protected/a.o -c -xc -fno-PIE -
+ cat
+ cc -shared -o /home/artur/Programming/C++/mold/fails/tmp/copyrel-protected/b.so -xc -
++ pwd
+ clang -fuse-ld=/home/artur/Programming/C++/mold/fails/../mold /home/artur/Programming/C++/mold/fails/tmp/copyrel-protected/a.o /home/artur/Programming/C++/mold/fails/tmp/copyrel-protected/b.so -o /home/artur/Programming/C++/mold/fails/tmp/copyrel-protected/exe
+ false

The resulting executable in ifunc-dso outputs nothing instead of Hello world.

~/Programming/C++/mold fix-tests-again !5 ?1 ❯ bash -x fails/ifunc-dso.sh        
+ set -e
++ dirname fails/ifunc-dso.sh
+ cd fails
++ basename -s .sh fails/ifunc-dso.sh
+ echo -n 'Testing ifunc-dso ... '
Testing ifunc-dso ... ++ pwd
++ basename -s .sh fails/ifunc-dso.sh
+ t=/home/artur/Programming/C++/mold/fails/tmp/ifunc-dso
+ mkdir -p /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso
+ cat
+ cc -o /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso/a.o -c -x assembler -
+ cat
+ cc -shared -fPIC -o /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso/b.so -x assembler -
++ pwd
+ clang -fuse-ld=/home/artur/Programming/C++/mold/fails/../mold -o /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso/exe /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso/a.o /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso/b.so
+ /home/artur/Programming/C++/mold/fails/tmp/ifunc-dso/exe
+ grep -q 'Hello world'

This can be caused by my /usr/bin/ld being a symlink to /usr/bin/mold
UPD confirmed that passing -fuse-ld=bfd to gcc & clang in this test fixes it.

~/Programming/C++/mold fix-tests-again !5 ?1 ❯ bash -x fails/run.sh      
+ set -e
++ dirname fails/run.sh
+ cd fails
++ basename -s .sh fails/run.sh
+ echo -n 'Testing run ... '
Testing run ... ++ pwd
++ basename -s .sh fails/run.sh
+ t=/home/artur/Programming/C++/mold/fails/tmp/run
+ mkdir -p /home/artur/Programming/C++/mold/fails/tmp/run
+ cat
+ cc -xc -c -o /home/artur/Programming/C++/mold/fails/tmp/run/a.o -
+ gcc -o /home/artur/Programming/C++/mold/fails/tmp/run/exe /home/artur/Programming/C++/mold/fails/tmp/run/a.o
+ readelf -p .comment /home/artur/Programming/C++/mold/fails/tmp/run/exe
+ grep -q mold /home/artur/Programming/C++/mold/fails/tmp/run/log
+ false

[rui314]

@Logarithmus Do you mind if I ask you to create a patch to fix issues you found? I can make a change myself but can't confirm that my change would fix the issue.

[Logarithmus]

@rui314, yeah, I'm currently working on it. I've installed Debian 11 inside lxd container. The idea is to compare generated binaries/logs between my system & the container.

[Logarithmus]

See #67 for follow-up on tests.

[rui314]

I cherrypicked bugfixes to https://github.com/rui314/mold/tree/release/0.9.2. Can you double-check that all tests pass now? If everything is fine, I'll tag it as v0.9.2.

[Logarithmus]

@rui314 I'm checking right now

[Logarithmus]

@rui314 2 small fixes needed:

1. #87
2. Bump version in Makefile (it's still 0.9.1 there).

[rui314]

@Logarithmus Thank you very much for testing! I cherry picked your change and bumped up the version number to 0.9.2. I'm tagging v0.9.2.

[rui314]

0.9.2 has been tagged.

[Logarithmus]

Thank you.