[Bug lto/65536] [5 regression] LTO line number information garbled

["manu at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65536

Manuel LÃpez-IbÃÃez <manu at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dodji at gcc dot gnu.org,
                   |                            |manu at gcc dot gnu.org

--- Comment #1 from Manuel LÃpez-IbÃÃez <manu at gcc dot gnu.org> ---
Even a large, but self-contained testcase would be helpful. Probably one can
figure out where the problem is by detecting that an overflow is about to
happen.

I don't understand how linemaps are supposed to work in LTO mode. Are they
exactly the same as without LTO and saved to the object file? Are they
recomputed? For all input files into the same line_table or one line_table for
each file separately?

Linemaps are designed to encode one single main file plus files included by
this main file. To encode multiple files, you will need multiple linemaps,
re-encoding everything into a single linemap seems fragile.

CCing Dodji, he may have a better idea where linemap may be failing.