[PATCH] D37791: [XRay][CodeGen] Use the current function symbol as the associated symbol for the instrumentation map

[David Blaikie via llvm-commits]

On Wed, Sep 13, 2017 at 11:46 PM Dean Michael Berris <dberris at google.com>
wrote:

> On 13 Sep 2017, at 23:20, Dean Michael Berris <dberris at google.com> wrote:
>
>
> On 13 Sep 2017, at 23:13, David Blaikie <dblaikie at gmail.com> wrote:
>
>
>
> On Wed, Sep 13, 2017 at 10:52 PM Dean Michael Berris via Phabricator <
> reviews at reviews.llvm.org> wrote:
>
>> dberris updated this revision to Diff 115167.
>> dberris edited the summary of this revision.
>> dberris added reviewers: dblaikie, echristo.
>> dberris added a comment.
>> Herald added a subscriber: JDevlieghere.
>>
>> Reworded description, added a test.
>>
>>
>> https://reviews.llvm.org/D37791
>>
>> Files:
>>   lib/CodeGen/AsmPrinter/AsmPrinter.cpp
>>   test/DebugInfo/X86/xray-split-dwarf-interaction.ll
>>
>>
>> Index: test/DebugInfo/X86/xray-split-dwarf-interaction.ll
>> ===================================================================
>> --- /dev/null
>> +++ test/DebugInfo/X86/xray-split-dwarf-interaction.ll
>> @@ -0,0 +1,76 @@
>> +; RUN: %llc_dwarf -split-dwarf-file=input.dwo -O3 -function-sections
>> -data-sections \
>> +; RUN:     -relocation-model=pic -filetype=obj -generate-type-units
>> -debug-compile -o %t %s
>>
>
> Have you tried this without fission/split-dwarf? I imagine it still
> reproduces (I think it's only type units that are important?)
>
>
>
> Not yet, but this was the case that caused the issues that I can
> reproduce. I can add another build here too.
>
>
> Actually, the "bug" wouldn't be fatal in that case, since we'd just have
> the wrong association. It's only when the debug section is stripped in case
> of split-dwarf that objcopy complains.
>

Right, but what I mean is there's a bug here (the wrong association) even
if it's not fatal. objcopy/split-dwarf exposed the underlying bug, but they
aren't integral to the bug, they exacerbate it.

Though perhaps I'm misunderstanding what the 'association' is and what it
being wrong might mean. Does that not result in any observable
problems/incorrect behavior in xray when its wrong like this?


> Landing this now, we can iterate more on it later.
>

Generally once a patch is sent for review it shouldn't be committed until
that review is approved. (the assumption being if something is sent for
review it's because review is needed)

- Dave


>
> Thanks for the reviews!
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170914/58cc257e/attachment.html>