https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99888
--- Comment #16 from Kewen Lin <linkw at gcc dot gnu.org> --- (In reply to Michael Matz from comment #15) > Umm. I just noticed this one as we now try to implement userspace live > patching > for ppc64le. The point of the "before" NOPs is (and always was) that they > are > completely out of the way of patchable but as-of-yet unpatched functions. > > For ppc that means the "before" and "after" NOPs cannot be consecutive. The > two > NOP sets being consecutive was never a design criteria or requirement. > > So, while the original bug is fixed by what was committed (local entry was > skipping the patching-nops), the chosen solution is exactly the wrong one :-/ Thanks for the input! Sigh, sorry that we picked up the wrong one :(, you may have noticed that the main consideration to choose the current one is to keep it align with the consecutive NOPs described by the documentation, we need a separate command line option as Segher's review comment in https://gcc.gnu.org/pipermail/gcc-patches/2022-August/600239.html. Now we have PR112980 filed for the requested behavior, let's discuss how to support it there.