From:	eric AT osiris DOT com DOT au (Eric Mills)
Subject:	Re: bash env parse broken
8 Sep 1997 22:50:24 -0700 :
Approved:	cygnus DOT gnu-win32 AT cygnus DOT com
Distribution:	cygnus
Message-ID:	<9709090132.AA10197.cygnus.gnu-win32@osiris.com.au>
X-Authentication-Warning:	sydney.dialix.com.au: osiruucp set sender to sydney!eric using -f
Original-To:	gnu-win32 AT cygnus DOT com
Original-Sender:	owner-gnu-win32 AT cygnus DOT com

B18 on NT4.0sp3 NTFS:

Thanks folks for the helpful hint re backslashes: all I can say is
"blind as a bat!" Sorry. I guess on seeing the equivalent of a 'core dump'
from cygwin32.dll I didn't look any further. There still must be a problem
that is not at all obvious with bash that caused the exception....

I would like to follow up on the second point re ^C and make. If you try the
following make file and hit ^C:

all:
	for x in *;\
	do sh -c echo;\
	done >/dev/null

then you will notice a few things:
1.  It never seems to work properly.
2.  More than 50% of the time you get fork_helper: child died message
3.  Less frequently one gets a cygwin_exception_handler message and everything
stops - some sort of deadlock situation - just keep killing bashes till
the world comes back.
4. If you run the same loop straight from bash you get an NT exception poppup
- not caught by cygwin_exception_handler! (or the child died message).

5. Lastly one notices just how SLOW this thing runs - on investigation I 
find that the power up time for bash is somewhere between 5 and 10 times longer
on NT + cygwin32 than UNIX for the identical hardware.
For us new NT users coming from the UNIX environment there should be a warning
statemant in the README file saying that shell script performance is much
less than one is accustomed to!

Would anyone care to comment on where the cycles are going?

Thanks, Eric Mills.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".

- Raw text -