Mail Archives: djgpp/2000/09/03/13:00:07
Message-ID: | <39B28CA5.14799CC8@user.rose.com>
|
From: | April <awhite AT user DOT rose DOT com>
|
X-Mailer: | Mozilla 4.7 [en]C-CCK-MCD {TLC;RETAIL} (Win95; U)
|
X-Accept-Language: | en
|
MIME-Version: | 1.0
|
Newsgroups: | comp.os.msdos.djgpp
|
Subject: | tracking an elusive crash
|
Lines: | 40
|
Date: | Sun, 03 Sep 2000 12:38:46 -0500
|
NNTP-Posting-Host: | 205.189.215.4
|
X-Trace: | client 967998751 205.189.215.4 (Sun, 03 Sep 2000 12:32:31 EDT)
|
NNTP-Posting-Date: | Sun, 03 Sep 2000 12:32:31 EDT
|
Path: | news.mv.net!newspeer.phoen-x.net!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.skycache.com!Cidera!portc01.blue.aol.com!peerfeed.news.psi.net!psinr!client!not-for-mail
|
Xref: | news.mv.net comp.os.msdos.djgpp:103134
|
I am working with btree code based on a text book. I get an elusive
crash that I cannot track down nor get any useful info post-crash.
djgpp gcc version 2.95.2 19991024
win95
The worst crash reboots the computer, the second worst gives a fatal
error 'you program has commited an illegal instruction'. Obviously in
both cases there is no call stack to symify.
I use a configuration file page size, and this is what makes the crash
elusive. If I use 512 bytes or 2048 bytes, the program can add about
17000 words to the tree (with duplicates ignored), and then delete 16999
(I leave one for testing), without a problem.
If I use a page size of 1024 (or just recently tried, 4096), I get the
fatal crash during the delete test, several 100 (maybe a 1000 or more)
words later.
I have tried to run the program in dos mode (ie. booted to dos), but the
GPF gives no callstack, and if I can believe the IP address, when I
locate it in the map file I finaly created, it is inside
libc.a(exceptn.o).
I have tried to increase my internal buffer that corresponds to the page
buffer, but too no avail; I tried to allocate larger memory blocks and
put an indicator before/after and check for those upon release, too no
avail. I tried to put an indicator at the start of all of my file
blocks and verify this upon any read process, to no avail.
HELP!
What can I try? I though it might be from redirecting a large file for
input and producing a large redirected stdout output, so I replaced the
input with a direct file read - it still crashed.
Suggestions?
April
- Raw text -