Many thanks! It worked. On Wed, 13 Nov 2024 at 16:33, LIU Hao via Gcc-help <gcc-help@xxxxxxxxxxx> wrote: > > 在 2024-11-13 20:46, Ahmad Nouralizadeh via Gcc-help 写道: > > I generated the assembly files. The relevant portion of the code > > (`text_segment_transformer.cpp:83`) is as follows: > > > > call syscall_hooker_cxx\n\t > > (Link: https://github.com/eunomia-bpf/bpftime/blob/808ce5990751af89f5af094c0e8a432b5f5aa35b/attach/text_segment_transformer/text_segment_transformer.cpp#L83) > > > > It has the same disassembly in both scenarios (`Ubuntu-18.04` and > > `Ubuntu-24.04`): > > > > call syscall_hooker_cxx > > > > But as mentioned in the previous threads, the final relocations have > > different types. Therefore, the assemblers seem to operate > > differently. Why does this happen? > > > > NOTE THAT, clang uses `callq` which results in relocatable symbols in > > both scenarios (although it seems to be call internally if I use > > `-masm=intel` option, link: > > I think this should be `call syscall_hooker_cxx@PLT`. > > It doesn't matter whether it's intel syntax or at&t; there's only ever one type of direct `call`. > > > -- > Best regards, > LIU Hao
