[llvm-bugs] [Bug 26519] Clang 3.8.0's "Target: powerpc-unknown-freebsd11.0" code generation is violating the SVR4 ABI (SEGV can result)

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=26519

Mark Millard <markmi at dsl-only.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |---

--- Comment #15 from Mark Millard <markmi at dsl-only.net> ---
(In reply to comment #14)

My interpretation of this is that the "lwz r30,24(r1)" is simply out
of sequence and should be later in the sequence:

> 0x0181b01c <__floatdidf+80>:	lwz     r30,24(r1)  <<<=== not here
> 0x0181b020 <__floatdidf+84>:	lfs     f2,0(r3)
> 0x0181b024 <__floatdidf+88>:	lwz     r3,-12(r30)
> 0x0181b028 <__floatdidf+92>:	lfs     f0,0(r6)
> 0x0181b02c <__floatdidf+96>:	lfs     f12,0(r3)
> 0x0181b030 <__floatdidf+100>:	fsub    f0,f1,f0
> 0x0181b034 <__floatdidf+104>:	fmul    f0,f0,f2
> 0x0181b038 <__floatdidf+108>:	fadd    f0,f0,f12
> 0x0181b03c <__floatdidf+112>:	fadd    f1,f13,f0
lwz     r30,24(r1)  <<<<<<<<<<=============== instead here
> 0x0181b040 <__floatdidf+116>:	addi    r1,r1,32
> 0x0181b044 <__floatdidf+120>:	mtlr    r0
> 0x0181b048 <__floatdidf+124>:	blr

(addresses not adjusted for the change in the above)

So that r30 would be restored between the main activity being
finished and r1 being adjusted (popping the stack frame). Of
course any place after the last local r30 use [after -12(r30)]
up to where I show it would also work.

I reported this issue as a comment on 26519 because I'm guessing
that the ordering problem is tied to trying to trying to avoid the
stack-handling ABI violations for FreeBSD: another incompleteness
in the coverage of the stack handling. If it is not that then a
separate submittal would likely be appropriate.

Since I'm allowed to I'm going to change the status to reopened so
that it does not look to have been fully fixed. Let me know if this
is inappropriate since I would not be the one fixing it.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20161127/694fab8b/attachment.html>