https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124836
--- Comment #2 from Tom Kacvinsky <tkacvins at gmail dot com> --- (In reply to Eric Botcazou from comment #1) > The commit is not supposed to have changed anything because OLD_MINGW had > been 0 for a while. Did you really experience any actual change here? I had been using GCC 12.1.0 (from https://github.com/msys2/mingw-packages) and this didn't exhibit any issues with path length. We have a test that opens a file with 259 characters in it. So, with the terminating \0, we were within the typical path length limitation of 260 for Windows. The commit in question was made after GCC 12.1.0 was released. And now I find with the GCC 15.1.0 I built (again with the MINGW-packages repo), the test bombed deep in the guts of the GNAT run time. The reason for that is now we exceed the limit as is set in adaint.c by four characters.
