[Webkit-unassigned] [Bug 223916] [JSC] Remove unnecessary copy constructor for ARMv7Assembler LinkRecord

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=223916

--- Comment #3 from Xan Lopez <xan.lopez at gmail.com> ---
(In reply to Darin Adler from comment #2)
> Comment on attachment 424622 [details]
> Remove user-provided copy-ctor, v1
> 
> If you are going to remove this then I think you need to do more:
> 
> 1) Research why it was added, since you are claiming it’s not needed. Let’s
> make sure we don’t lose the context of why someone thought they needed it.
> Verify they were mistaken or that something has changed.

Thank you, this was actually the right thing to do first. This was introduced way back in 2012 as an optimization, the claim being it's a significant speed-up over the compiler-provided default. See https://bugs.webkit.org/show_bug.cgi?id=90930

I think this may very well be still the case, perhaps I should try to measure it again, but then I'd say that we at least:

- Should add a comment about it, with a link to the bug.
- Find a way to shut up GCC about it, maybe by doing what you suggest in the last comment.
- If we leave the operator in place, maybe we might want to fix the operator (at least) in ARMv7, since it's using a 'void' return type instead of the more common/correct reference type? The version in the ARM64 assembler file is fine.

> 
> 2) Remove the unused CopyTypes type and copyTypes data member and remove the
> use of the union entirely since this is the only place it was used.
> 
> If you want to take the GCC warning literally, you could add a copy
> constructor that uses copyTypes instead of removing the assignment operator.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20210330/34f70e1c/attachment.htm>