cvs commit: doc/en_US.ISO8859-1/books/faq book.sgml

Giorgos Keramidas keramida at
Sat Jul 5 14:55:39 UTC 2008

On Sat, 05 Jul 2008 09:12:46 +0200, Gabor PALI <pgj at> wrote:
>> The suggestion about STRIP= and CFLAGS+='-g' is a bit old now too.
> I had the same suspicion (especially for STRIP), but it seems it managed
> to escape our (my, gabor's and the reviewers') observation somehow.
>> cd /usr/src/usr.sbin/ppp env DEBUG_FLAGS='-g' make clean env
>> DEBUG_FLAGS='-g' make install
> However, I think CFLAGS still can be used for inclusion of debug
> symbols, although your env(1) based solution is a good (and very
> elegant!) idea for ones who do not want to modify their /etc/make.conf
> every time.

CFLAGS with optimizations may cause 'bogus' tracebacks, because of
function inlining or other optimization changes.  It's usually a good
idea to disable all optimizations when tracking down bugs with gdb(1),
but that may be risky too.  Some bugs only show when optimizations *are*
enabled, and may turn to "Heisenbugs" when optimizations are disabled.

Having said that, DEBUG_FLAGS doesn't override CFLAGS.  It merely makes
sure that:

  * CFLAGS includes the value of DEBUG_FLAGS.

  * No strip(1) run is done to remove debugging symbols.

>> Does the following patch for this question look ok to you?
> Your version is very nice, I would be happy to add it to the FAQ.  I
> would like to ask doc committers (and FAQ patchletter reviewers) reading
> this mail to review and comment on Giorgos' patch to have a third-party
> review.
> Giorgos, would you approve my commit for this patch in advance?  Sorry
> for that, but I am actually a guest here :)

Of course.  Please feel free to commit it either "as is" or in a
modified form.

> PS: Hey, Giorgios, what about my checkupdate patches?  Have you tried or
> reviewed them?  I already started to use them as you will see very soon :)

I just started getting "back" from ${real_life}.  I'm slowly catching up
with email, but I will try to reply soon-ish about those.

Thanks :)

More information about the cvs-doc mailing list