Using -gstabs (or -gstabs+) in a general purpose Makefile is probably
a bad idea because it may force gcc to use stabs; stabs might not be
the best debugging format available on a user's system. For example,
DWARF is supposed to be vastly superior all around from what I hear.
Apple just adopted DWARF in Xcode 2.3 about 4 months ago and I'm
already reading many anecdotal stories about how DWARF has fixed all
the problems people had with stabs. I know for a fact that using DWARF
instead of stabs cut the OSG debug files binary size by about 90% (1GB
to 100MB). I'm thinking about making this the default option for the
Xcode projects.

Also, for -gstabs+, if the flag is accepted and you are using debug
tools other than raw gdb, I think the tools may crash or be unable to
read the program.

I think -ggdb picks a default, though it is not necessilariy the best
option. (For example, I think Apple still picks stabs as the default.)
But this is probably the safest option to use in a general purpose
Makefile. I'm not sure what gets picked if you include both -ggdb and
another flag like -gstabs, but it's probably best to avoid putting
both down.

-Eric




Hi Mark,

I put in the gstabs+ argument quite a while back when I posted a problem on
the gcc mailing list and was recommended this option.  I now read the doc on
it and it seems like it is an option specific to BSD, so we should change
makedefs accordingly.

However, I would not expect changes destined for BSD platforms to affect
other platforms, so it would seem that this is a bug.  BTW, we use both
-gstabs+ and -ggdb currently.

-don

On 9/22/06, Mark Acosta <[EMAIL PROTECTED]> wrote:
>
>  Hi guys,
>
>     I just put together a brand spanking new linux box running 64 bit
> fedora core 5 and much to my suprise, I found that gdb was unable to
access
> any memory while debugging an osg program. After much head banging, I
> tracked the problem down to the -gstabs+ compile option. I switched to
-ggdb
> in osg, Producer and OpenThreads and in my own code and now gdb works
fine.
> I was running 32 bit fedora core 5 before and there wasn't any problem
with
> gdb there. I'm not sure if this is a 64 bit gdb or gcc/g++ bug or what.
> Anybody else ever run into this problem or even not have this problem? By
> the way, I'm using gcc 4.1.1 and gdb 6.3.
>
>     Also, I noticed that when doing a "make instlinks" , osg will install
> in /usr/local/lib64, which I assume is correct, but Producer and
OpenThreads
> will always install in /usr/local/lib even if I specify ARCH=64. I took a
> look at the make files but I don't see anything obviously wrong.
>
> Thanks,
>
> Mark Acosta
>
> _______________________________________________
> osg-users mailing list
> [email protected]
> http://openscenegraph.net/mailman/listinfo/osg-users
> http://www.openscenegraph.org/
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://openscenegraph.net/pipermail/osg-users/attachments/20060924/ffaca0af/attachment-0001.html

----------------
_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/

Reply via email to