httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Cliff Woolley <>
Subject old_write filter problems
Date Tue, 19 Mar 2002 07:18:13 GMT

I've been trying to debug a SEGV I'm seeing in httpd-pop3 at seemingly
random times (I can reproduce it kind of, but I don't have the exact
circumstances nailed down)... but frankly the bug could just as easily be
in the old_write filter as in pop3 at this point (and I'm leaning toward
blaming old_write at the moment).  Here's as much of the backtrace as is
useful (the rest varies):

(gdb) bt
#0  apr_brigade_write (b=0xa003420, flush=0x80c9cf0 <ap_filter_flush>,
    ctx=0x81f85c0, str=0xbfffd57c "+OK message 4 deleted\r\n", nbyte=23)
    at apr_brigade.c:404
#1  0x80cbed1 in buffer_output (r=0x8205270,
    str=0xbfffd57c "+OK message 4 deleted\r\n", len=23) at protocol.c:1376

0xa003420 is not actually the address of a brigade, so of course
apr_brigade_write() SEGV's.  That address comes from buffer_output() which
looks up the old_write_filter_ctx->bb (around line 1370 of protocol.c),
which is non-NULL but invalid (0xa003420 in this case, but sometimes a
value like 0xa).

Anybody seen anything like this?


   Cliff Woolley
   Charlottesville, VA

View raw message