Fix Issue 20318 - Illegal instruction (core dumped)

[RazvanN7]

void main()
{
    throw new Exception("msg");
}

When an exception is thrown, typically it is allocated and managed by the garbage collector: the compiler replaces the expression with a call to a druntime hook, in this situation _d_newclass. If the code is compiled with profile=gc the hook is further replaced with a call to another hook d_newclass_trace that collects some stats and later calls _d_newclass.

If the above code is compiled with -dip1008 the exception no longer uses the gc to be managed; it is manually managed in druntime via reference counting. So if -dip1008 and -profile=gc are used together the following thing happens: (1) the exception throwing line is rewritten to a call to _d_newthrow in druntime which does the refcount initialization for the exception then (2) toTraceGC in dmd is called that replaces any allocation hooks with trace allocation hooks. Since _d_newthrow does not do any gc allocations and therefore does not have an associated trace function, toTraceGC gets confused and asserts.

The fix is to simply return from toTraceGC if the hook is not a gc allocation hook.

[dlang-bot]

Thanks for your pull request and interest in making D better, @RazvanN7! We are looking forward to reviewing it, and you should be hearing from a maintainer soon.
Please verify that your PR follows this checklist:

• My PR is fully covered with tests (you can see the coverage diff by visiting the details link of the codecov check)
• My PR is as minimal as possible (smaller, focused PRs are easier to review than big ones)
• I have provided a detailed rationale explaining my changes
• New or modified functions have Ddoc comments (with Params: and Returns:)

Please see CONTRIBUTING.md for more information.

---

If you have addressed all reviews or aren't sure how to proceed, don't hesitate to ping us with a simple comment.

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	20318	normal	Illegal instruction (core dumped)

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub fetch digger
dub run digger -- build "master + dmd#10517"

[codecov]

Codecov Report

Merging #10517 into master will decrease coverage by <.01%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master   #10517      +/-   ##
==========================================
- Coverage    1.81%    1.81%   -0.01%     
==========================================
  Files         148      148              
  Lines       77438    77440       +2     
==========================================
  Hits         1405     1405              
- Misses      76033    76035       +2

Impacted Files	Coverage Δ
src/dmd/e2ir.d	0% <0%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 83fde70...eae1cd8. Read the comment docs.

[Geod24]

Doesn't look like a compilable test is what's needed there, but rather a runnable.

[RazvanN7]

I don't get it. This was an ICE that was triggered without my patch by the testcase that I added. How can it not be covered?

[wilzbach]

I don't get it. This was an ICE that was triggered without my patch by the testcase that I added. How can it not be covered?

It looks like someone destroyed the coverage:

https://codecov.io/gh/dlang/dmd/commits

It's on 1.8%
This commit caused it: https://codecov.io/gh/dlang/dmd/commit/6bb95b5bed514070e13ec9a497f47a80e8dfc5c7
#10446

Misread the issue

[marler8997]

code coverage issue fixed here: #10520

[RazvanN7]

Is there any way to restart codecov?

[wilzbach]

Yes, you need to restart the respective build (CircleCi). I just did so for you.

[WalterBright]

Nice work!

I do like your explanation in the opening message here. Can you boil it down to its essence and add it as a comment to the code? Because it isn't obvious.

[RazvanN7]

Nice work!

Thank you!

I do like your explanation in the opening message here. Can you boil it down to its essence and add it as a comment to the code? Because it isn't obvious.

Done.

[PetarKirov]

Wouldn't it be better to make the call to toTraceGC conditional on whether the NexExp field ne.thrownew is true here:

dmd/src/dmd/e2ir.d

Lines 1676 to 1678 in a7017a9

	const rtl = global.params.ehnogc && ne.thrownew ? RTLSYM_NEWTHROW : RTLSYM_NEWCLASS;
	ex = el_bin(OPcall,TYnptr,el_var(getRtlsym(rtl)),el_ptr(csym));
	toTraceGC(irs, ex, ne.loc);

And then assert that s != getRtlsym(RTLSYM_NEWTHROW) inside toTraceGC?

[RazvanN7]

@PetarKirov That was my first thought also, but the call to toTraceGC is made regardless of whether profile=gc is set or not and the check for it is done inside the function, so I interpreted it as a means to keep the main code clean.

[PetarKirov]

Fair enough