[PATCH] D134097: [NFC] Refactor creation of symbol+addend references
Maksim Panchenko via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Oct 7 15:21:51 PDT 2022
maksfb accepted this revision.
maksfb added a comment.
Add "[BOLT]" to the title. Replace "this code" in the Summary with the description of the code. Otherwise LGTM.
================
Comment at: bolt/lib/Core/BinaryFunction.cpp:1054
+ (void)ReplaceSuccess;
+ assert(ReplaceSuccess && "failed to replace mem operand with symbol+off");
}
----------------
rafauler wrote:
> yota9 wrote:
> > Maybe change assert to if + exit ?
> Are we migrating away from assertions? This is not supposed to fail at all, unless we modify our codebase in weird ways, so I would like to assert here. But I'm not sure how hard are people trying to get rid of assertions.
>
> Another reason this might fail is if somebody is writing a new backend and fails to replace the memory operand of a given instruction that they need to support replacing.
It it's a validation of internal assumptions, not an assumption about the input, then it should be an assertion.
================
Comment at: bolt/lib/Core/BinaryFunction.cpp:1054
+ (void)ReplaceSuccess;
+ assert(ReplaceSuccess && "failed to replace mem operand with symbol+off");
}
----------------
maksfb wrote:
> rafauler wrote:
> > yota9 wrote:
> > > Maybe change assert to if + exit ?
> > Are we migrating away from assertions? This is not supposed to fail at all, unless we modify our codebase in weird ways, so I would like to assert here. But I'm not sure how hard are people trying to get rid of assertions.
> >
> > Another reason this might fail is if somebody is writing a new backend and fails to replace the memory operand of a given instruction that they need to support replacing.
> It it's a validation of internal assumptions, not an assumption about the input, then it should be an assertion.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D134097/new/
https://reviews.llvm.org/D134097
More information about the llvm-commits
mailing list