hi, I've been testing b-f on PReP for a while, and encountered strange crashes (or restarts) of dbootstrap at the end or after the debootstrap step -- this is related to the serial console install problems related in #126370. now, the conclusion I've come to is that by disabling line 89 in debootstrap's "woody" shell script, which reads 87: cat <$TARGET/dev/initctl >/dev/null & 88: CAT_PID=$! 89: on_exit "kill $CAT_PID >/dev/null 2>&1" dbootstrap has a chance to survive the end of debootstrap; looks like that kill, which seems to just show up as "NNNN: no such pid" when you look closely, somehow manages to take dbootstrap with it. my manual tests have shown that dbootstrap's exit code in this case is 143, which I also can't explain: apart from some "exit(1)" statements, the only one which sets such a high value is the exit(128+sig) at crash.c:45, but that is in a signal handler which only gets installed for SEGV, as far as I follow the code. 143 is 128 + SIGTERM. well, I'm a bit confused -- the only clear thing is that with that line commented out, stuff which occurs in dbootstrap after debootstrap is finished does happen, while without that change it crashes reproducibly (and I've done quite a few test runs this night). I hope somebody can solve that mystery... ciao, cm. -- ** christian mock in vienna, austria -- http://www.tahina.priv.at/~cm/ Rechtlich gesehen, besteht kein Anspruch auf Weiterleitung von Artikeln, die Bilder von Führerscheinumschlägen enthalten. ghw in at.usenet
Attachment:
pgpvbq5g4GIzO.pgp
Description: PGP signature