[2181] in SIPB_Linux_Development
Re: Official policy on what build flags to use for the release?
daemon@ATHENA.MIT.EDU (Theodore Y. Ts'o)
Wed Sep 30 12:25:58 1998
Date: Wed, 30 Sep 1998 12:25:42 -0400
From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: Greg Hudson <ghudson@MIT.EDU>
Cc: Salvatore Valente <svalente@MIT.EDU>, linux-dev@MIT.EDU
In-Reply-To: Greg Hudson's message of Wed, 30 Sep 1998 08:44:07 EDT,
<199809301244.IAA00306@nephthys.grey17.org>
The only thing I would add to Greg's excellent discription is that at
least with gcc, there is no difference between a stripped binary which
had been compiled with -g, and a stripped binary which was compiled
without -g. -g only adds debugging information, but it doesn't actually
change the code generation. This is not necessarily true with all
compilers, but it is true with gcc. (Some compilers generate slightly
less efficient code if you specify -g, and some compilers do not permit
the combination of -g and -O.)
So the reason why I suggested "-g -O" while installing stripped binaries
is that we would be installing an efficient (in space and time) binary
on the packs for users to use, but in the case where we want to debug a
core dump, we would be able to go to the archived unstripped binary, and
in combination with the source snapshot used to produce that binary, we
would be able to do pretty much full-scale debugging on the core dump.
- Ted
