[llvm-jitlink] Replace IR backtrace symbolication testcase with asm.

[lhames]

Some builds (e.g. https://lab.llvm.org/buildbot/#/builders/154/builds/26180) failed for this testcase, for reasons that aren't immediately clear from the logs.

Since the options being tested are for debugging only, this commit switches to testing Darwin/arm64 only.

Whne similar backtrace symbolication options are added to lli we can reintroduce a generic testcase and start investigating these failures.

[dyung]

Hi @lhames, the test still seems to be intermittently failing on my AArch64 darwin bot, not sure if you have any ideas what might be happening: https://lab.llvm.org/buildbot/#/builders/190/builds/33941

[mysterymath]

Hi @lhames, the test still seems to be intermittently failing on my AArch64 darwin bot, not sure if you have any ideas what might be happening: https://lab.llvm.org/buildbot/#/builders/190/builds/33941

This is happening in Fuchsia too; accordingly, I'm going to issue a revert.

See https://luci-milo.appspot.com/ui/p/fuchsia/builders/toolchain.ci/clang-host-mac-x64/b8693176322074417553/overview

[boomanaiden154]

Given that revert just reverts the fix forward which exposes more breakage from the underlying commit, I've also reverted the original commit in the mean time. c11df52

[mysterymath]

Given that revert just reverts the fix forward which exposes more breakage from the underlying commit, I've also reverted the original commit in the mean time. c11df52

Yikes, sorry about that.

[lhames]

Hi @lhames, the test still seems to be intermittently failing on my AArch64 darwin bot, not sure if you have any ideas what might be happening: https://lab.llvm.org/buildbot/#/builders/190/builds/33941

This is happening in Fuchsia too; accordingly, I'm going to issue a revert.

See https://luci-milo.appspot.com/ui/p/fuchsia/builders/toolchain.ci/clang-host-mac-x64/b8693176322074417553/overview

Hi @mysterymath, @boomanaiden154 -- The test contains '# REQUIRES: system-darwin'. Is system-darwin being set on Fuchsia? That seems like a bug.

[lhames]

Oh, that's a Darwin build of a Fuchsia toolchain I guess. It's also x86-64, so definitely shouldn't be running this test at all.

I'm going to re-land the original patch without a test (it's best-effort diagnostic only), then re-land the test with a target-aarch64 condition added.

@dyung -- I would still expect the test to run on your tester even with this modified config. Is there any chance you could run it manually and share the full backtrace? That'll help identify what's going wrong with the symbolication. (I suspect the test is getting killed, so the backtrace doesn't include the expected frame. I'm not sure what would be killing it though)

[lhames]

Symbolication support and options re-landed in #175469

Testcase re-land in #175476

[lhames]

@dyung #175537 adds a new diagnostic that should help to figure out what's going on with that failure. I'll land tomorrow morning Sydney time, or you're welcome to merge before then if you'd like. I'm hoping to get at least one run in with that diagnostic applied to see if we can figure out where the wild address for main is coming from.

Once we've got that info the testcase can be disabled or removed for now.

[dyung]

@dyung #175537 adds a new diagnostic that should help to figure out what's going on with that failure. I'll land tomorrow morning Sydney time, or you're welcome to merge before then if you'd like. I'm hoping to get at least one run in with that diagnostic applied to see if we can figure out where the wild address for main is coming from.

Once we've got that info the testcase can be disabled or removed for now.

Thanks. I've got a lot of meetings today, but I'll try to get to it today and see if I can figure out why it is failing.