DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG 
RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT
<http://nagoya.apache.org/bugzilla/show_bug.cgi?id=18756>.
ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND 
INSERTED IN THE BUG DATABASE.

http://nagoya.apache.org/bugzilla/show_bug.cgi?id=18756

Apache core dump when using LDAP authentification.





------- Additional Comments From [EMAIL PROTECTED]  2004-02-19 02:47 -------
This appears to be triggering a gcc on Sparc bug -- read on...

Note the backtrace given above was hard to obtain, e.g. I could only obtain it 
when *not* using maintainer mode on configure, etc.

Once I put breakpoints in the routine in question in each child process I 
could single step through the assembly (for some reason I did not have source-
level information even when building with maintainer mode enabled and thus -g -
- pointers as to how this might happen would be appreciated) until it blew up.

The offending assembly instruction is 'std' (this is on Sparc Solaris 8) and 
the error is an invalid alignment error.  This smelled of a compiler bug -- 
and sure enough google has lots of claims of various versions of gcc making 8-
byte alignment assumptions when it can't legitimately do so -- and then using 
ldd and std instructions which require such alignment.  Thus seeing this only 
as assembly was not a bad thing -- except that I need to carefully verify 
which struct field its blowing up on...

Does anyone know of a gcc version that is known to have this error fixed?  
Alternatively, does someone know of flags to pass gcc to work around this?  
Note this error appears to be specific to -O2 optimization (which I removed 
from my configure env vars, but I can't seem to get rid of it).  Ideally the 
source for this module or the gcc compilation options for just it (not all of 
Apache) could be adjusted to skirt this issue.

I'll be playing with the possibilities here...

[I'm currently still using the ancient gcc 2.95.2 as it's what's already 
installed on the machine.]

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to