strip-flags

[BlackPhoenix]

Ok...before anyone continue reading...keep in mind I am one of the "users" of gentoo...and haven't been paying much attention on the dev part until very very recently...so if Im wrong on anything, just point it out

I just noticed a few days back how, during compiling, my optimisation flags were not being used in many builds... Thinking I might have commented lines by mistake in make.conf, I'd just try and emerge something else, and then they were used...so I figured it was only my imagination, or something I didnt understand...until I looked at an ebuild and saw this...

*warning, personal opinion, and maybe not actual fact, follows*

I've started using Gentoo because, even though Im not very knowledgeable about Linux (I jumped from Lycoris Desktop/LX to Gentoo, if that gives you guys an idea...), but I was just drooling at the idea of near total control over what my machine did...

Now, here is the deal...I can understand when some flags that are known to, almost 100% surely break a package, are filtered...

But stripping the flags almost systematicaly because some users don't know what they are doing? Wouldn't that be something the user needs to choose, not the ebuild's creator? Which was, afterall, the whole point in having make.conf CFLAGS and CXXFLAGS variables in it...? No?

So if this is the case...why are the flags being stripped without even telling me, really? Had to go read ebuild for that...

Now if it was only one, cool...but with this all over the place, Im gonna have to go double check all the ebuilds I feel like emerging, and possibly editing them, and that, everytime I rsync...? Kindda awkward, and not truly with what I got from Gentoo's phylosophy...

So wouldn't it be a better idea to have a variable in make.conf, something like ALLOW_FLAG_STRIPPING = 0 if you know what you are doing and don't feel like having your flags stripped down? (The ones I used were perfectly stable, and never broke any package, even when I edited the ebuild to remove strip-flags, but did make stuff run kindda faster...so, well...)

------------------ end of rant ------------

So well...if I just didnt get something (im wrong on what strip-flags does, or I missed the track completly), feel free to correct me. Otherwise...am I the only one who think like this?

-Frankie

[rac]

I have basically the same understanding as you do as to how strip-flags works. It seems to allow an optimization level (-Osomething), an -mcpu=something, an -march=something, and -pipe. Everything else gets dumped.

The first time I upgraded my athlon from GCC 3.1 to GCC 3.2, I had lots of stability problems with -O3. Mozilla would take off and peg the CPU about three times an hour, XEmacs would do the same sometimes and would simply segfault others. X Errors were also prevalent, taking down terminals and applications. So I have moved all of my machines that use GCC 3.2 to -O2. I have had first-hand experience with what appears to me to be silent bad code generation by GCC 3.2 when using certain "optimizations".

I tell this story to explain the frame of mind I was in when I first read this comment: