On Feb 4 00:04, Takashi Yano wrote: > On Sat, 3 Feb 2024 15:27:06 +0100 (CET) > Johannes Schindelin wrote: > > On IRC, you reported that the thread would crash if `cons` was not fixed > > up. The symptom was that that crash would apparently prevent the exit code > > from being read, and it would be left at 0, indicating potentially > > incorrectly that the non-Cygwin process succeeded. > > > > I wonder: What would it take to change this logic so that the crash would > > be detected (and not be misinterpreted as exit code 0)? > > I am not sure, but I think it is necessary to modify: > pinfo::exit() > pinfo::meybe_set_exit_code_from_windows() > pinfo::set_exit_code() > > I guess detecting crash of sbub process needs modification of > spawn.cc.
Dumb question: If, as Johannes said, the error code cannot be fetched, can't we set the error code to a POSIX return code indicating a signal? I.e., checking for WIFSIGNALED() returns 1 and WTERMSIG() returns, say, SIGKILL or something? Corinna
