Mail Archives: cygwin-developers/1998/08/02/05:27:46
After further investigation, I have determined that a programming error on
my part was responsible for bash not understanding when a command was
finished. My not-so-buggy modified bash no longer causes any crashes. ash
is still slightyly faster for configures than my bash. :( The result on my
machine is that the modified bash performs a configure 10% faster than the
unmodified bash 2.02.1.
I am still trying to track down the conditions under which the hang was
occuring. The nature of the bug that I had would mean that bash was not
calling waitpid for the spawned processes. I suspect that there might be a
race condition in the signal handling code, but there is still more
detective work to be done.
I rebuilt the cygwin dll with -DDEBUGGING. The slightly more informative
messages that I receive are:
(D:\CYGNUS\B19\H-I386-CYGWIN32\BIN\SH.EXE 2495) block_sig_dispatch: [main]
WFSO(
sig_dispatch<0x2C>) failed, error = 0
(D:\CYGNUS\B19\H-I386-CYGWIN32\BIN\SH.EXE 2495) block_sig_dispatch: [main]
WFSO(
sig_dispatch<0x2C>) failed, error = 0
(D:\CYGNUS\B19\H-I386-CYGWIN32\BIN\SH.EXE 2495) lock_cs: WFSO(cs), rc 258,
error
288
- Raw text -