https://sourceware.org/bugzilla/show_bug.cgi?id=27491
Alan Modra <amodra at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Ever confirmed|0 |1
Last reconfirmed| |2021-03-02
--- Comment #1 from Alan Modra <amodra at gmail dot com> ---
powerpc64le on a similar testcase with
_start:
pld 3,__start_xx@got@pcrel
pld 4,__stop_xx@got@pcrel
does a similar optimisation without complaint, generating a binary that
performs two pla instructions setting both registers to zero pc-relative (which
isn't zero when the shared library is loaded somewhere other than address
zero). Technically that is wrong since the value of an undefined weak symbol
is zero, and so the got entries should be left in order to load an absolute
zero. But it might be reasonable in this case since you expect __start_xx and
__stop_xx to have an address somewhere in the executable. However, I think
I'll fix this.
In contrast
.weak xxx
.global _start
_start:
pld 3,xxx@got@pcrel
leaves the pld alone when xxx is undefined, loading the reg with zero from the
got entry.
--
You are receiving this mail because:
You are on the CC list for the bug.
