[Bug ld/26312] ld produces broken PLT on aarch64 with BTI+PAC

[nsz at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=26312

--- Comment #6 from Szabolcs Nagy <nsz at gcc dot gnu.org> ---
note on 32bit arm, sh and or1k .plt sh_entsize == 4 which
has nothing to do whith the plt entry size (which can
vary on arm). on sparc the logic is more complicated and
the plt entry size is used in some cases but the plt
header is bigger (it's a multiple of the entry size though).
and there are targets where it is 0.

to me this tells that sh_entsize of .plt is not reliable
and should be just ignored by tools (ltrace cannot use
it either).

i will set it to 0 in all cases on aarch64 and hope
nothing will break.

(note that the bug only affects executables, not shared
libs: only executables use special plt for bti because
then the plt can provide the canonical address of function
symbols, in libraries the plt is not called indirectly)

-- 
You are receiving this mail because:
You are on the CC list for the bug.