https://sourceware.org/bugzilla/show_bug.cgi?id=32961
--- Comment #22 from H.J. Lu <hjl.tools at gmail dot com> --- (In reply to Zhiyuan Lv from comment #21) > (In reply to H.J. Lu from comment #13) > > (In reply to Zhiyuan Lv from comment #12) > > > Hi H.J., > > > > > > On Tue, May 13, 2025 at 07:15:14AM +0000, hjl.tools at gmail dot com > > > wrote: > > > > https://sourceware.org/bugzilla/show_bug.cgi?id=32961 > > > > > > > > --- Comment #8 from H.J. Lu <hjl.tools at gmail dot com> --- > > > > (In reply to Zhiyuan Lv from comment #7) > > > > > And, the GCC version I used is 14.2.0, which will generate per > > > > > function > > > > > section of "__patchable_function_entries". Thanks! > > > > > > > > Now, the question is if GCC should generate a unique section name for > > > > __patchable_function_entries -ffunction-sections. If I give you a > > > > > > Just curious: is the same name a problem here? I saw that gcc 14 > > > generates __patchable** sections for each function with the same name, > > > and they seem to work fine: each section has its own relocation section > > > respectively. Why does "-ffunction-sections" cause problems? Thanks! > > > > > > > It is the similar problem on GCC side. -fpatchable-function-entry doesn't > > work with -ffunction-sections. My GCC patch should fix it. > > Could you elaborate more why the two parameters cannot work together? They work. > I am still wondering whether it is possible to avoid compiler change. With > GCC 14 in Ubuntu 24.04, I can see that gcc generates > __patchable_function_entries sections per function with the same name, and > GAS/LD seems to treat them separately so they do not introduce unnecessary > dependencies. So, the same name of __patchable_function_entries is not a > problem. > > IIUC, all we need is unique name for .pushsection, and your latest patch has > already ensured that. Today I also tested that with unmodified gcc 14, all > my tests passed. Thanks! Great. -- You are receiving this mail because: You are on the CC list for the bug.
