This is the mail archive of the
cygwin@sourceware.cygnus.com
mailing list for the Cygwin project.
RE: Weird behavior when using "-I" with gcc-2.95.2
- To: "'Mumit Khan'" <khan at NanoTech dot Wisc dot EDU>, Fontenot Larry A DLVA <FontenotLA at NSWC dot NAVY dot MIL>
- Subject: RE: Weird behavior when using "-I" with gcc-2.95.2
- From: Fontenot Larry A DLVA <FontenotLA at NSWC dot NAVY dot MIL>
- Date: Thu, 27 Apr 2000 16:36:51 -0400
- Cc: "'cygwin at sourceware dot cygnus dot com'" <cygwin at sourceware dot cygnus dot com>
Here are the strace outputs. h_strace.txt was compiled on my h:\ drive and
it worked. i_strace.txt was compiled on the UNIX via SunLink server and it
failed. Note line numbers i_strace.txt:1114 and h_strace.txt:1124. This is
where the first discrepancy is between the two runs. It appears the errno
is BAD_PATHNAME, in the broken version, versus PATH_NOT_FOUND, in the
working version.
By the way, I downloaded a trial version of Omni NFS Client Lite, and
mounted the I:\ drive using NFS instead. That also clears the problem right
up.
Thanks,
Larry
<<strace.tar.gz>>
-----Original Message-----
From: Mumit Khan [SMTP:khan@NanoTech.Wisc.EDU]
Sent: Thursday, April 27, 2000 2:10 PM
To: Fontenot Larry A DLVA
Cc: 'cygwin@sourceware.cygnus.com'
Subject: RE: Weird behavior when using "-I" with gcc-2.95.2
On Thu, 27 Apr 2000, Fontenot Larry A DLVA wrote:
>
> Well that was very interesting. Everything compiles just fine on
the local
> hard disk when using the "-I ./". So then I tried my H:\ drive.
That is a
> drive I have mapped from a WinNT server. Once again, everything
works just
> fine. So then I went to my I:\ drive, which is a Sun Solaris file
server
> using SunLink software to emulate a WinNT server. That is when I
start
> having problems. Apparently, gcc is seeing something differently
when
> compiling on my I:\ drive. Should I try to use strace at this
point to find
> out what the difference might be?
Normally I'd blame inode generation on remote drives, but that's not
it
since I've turned off inode based path lookup optimization in gcc.
It does look like a problem with Cygwin interaction with SunLink (I
know
nothing of this software). Now that we know where the problem lies,
we
should try to isolate it by writing the shortest possible testcase.
Maybe
a problem with `stat' on SunLink server?
I suppose the strace output may be a good place to start. It's going
to be
large, so do compress the output using gzip before posting.
Hopefully the
Cygwin gurus can take over and solve this mystery.
I'm gone till early next week, so can't be of any help till then.
Regards,
Mumit
strace.tar.gz
--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com