As many other packages have this useflag, I will simply give my result. standard ebuild (lyx 1.5.4): ${S} size at the end of src_install - 443065538b genlop -t: 59 minutes and 45 seconds --disable-pch --without-included-boost: ${S} size at the end of src_install - 108779392b genlop -t: 32 minutes and 16 seconds Of course, my use of ccache may have had some impact on the time, but not to this level And about boost: I asked about using system boost and last time it was rejected, but I'm not sure whether it was a real issue or ebuild maintainer dodging I needed a little patch, one part of it I'm almost sure is correct (limits), the other one I'm not sure, but it seems to work. This patch was probably needed, cause I'm using gcc 4.3.0 and it was triggered by pch, not boost.
Created attachment 147160 [details, diff] patch to build with --disable-pch
Forgot to add: dev-libs/boost-1.34.1-r2
hi, fyi: - pch are should be disabled by default from upstream in the forthcomming 1.6. - the ebuild i have prepared for 1.6 have also boost flag. - wrt compilation time monolithic flag should be working in 1.6 too to the attachments - if you are sure --disable-pch && gcc 4.3.0 cause compilation to fail it would be good to come in lyx devel list with your patches if no from other reasons than to review. "I'm not sure, it seems to work" patches are imho not good idea to put directly into portage.
Well, the problem with this patch is: while #include <limits> seem to match some of the error I can google, abort() thing is a guess, cause I'm not sure why it fails in the first place, as <stdlib.h> is included and std::abort() didn't work.
this is not the place where to discuss these patches.
should be fixed in 1.5.5 http://www.lyx.org/trac/changeset/24069
*** Bug 220807 has been marked as a duplicate of this bug. ***
1.5.5 in portage
And still no pch USE :) Can we please, please have pch as an option, and also have it use system boost? I don't like having to keep a copy of release LyX in my overlay just for this :P
i promised these for 1.6 series. as far as 1.5 concerned i prefer stability over optimalization tweaking.
Well, I've just tested it (1.5.5) and it's still broken with --disable-pch --without-included-boost. Compilation fails at src/support/abort.cpp again. However the fix for it is a bit silly. All that needs to be done is move the block #include <cstdlib> #ifdef HAVE_STDLIB_H # include <stdlib.h> #endif outside of namespace lyx {...}
:D at least you see why i'm hesitant to change these things.
But it seems it was a true lyx bug. If I understand correctly what pch.h did, it was including all those c++ headers at global scope, when it was taken out, those includes in abort.cpp were simply at the wrong place to work.
I just noticed that the latest note on 1.5.X svn branch is : uwestoehr: installer: sync with new version from trunk and prepare for LyX 1.5.6 but this bug is still not fixed. Any chances on getting the fixed I proposed (moving the includes outside namespace lyx{}) into 1.5.6 ?
is it better with lyx 1.6.0_beta3 ?
It's hard to tell, as I haven't tested 1.6.0beta. I tend to stay away from betas, unless there's a very good reason not to, like there's no such thing as a stable version of the package or stable is so far behind svn, that 100% svn is a better route. And anyway, I don't see that file in 1.6.0 svn tree.
(In reply to comment #15) > is it better with lyx 1.6.0_beta3 ? > I've been using a live svn ebuild i made for the 1.6 branch for a while now - all combinations of pch, boost and monolithic (think kdeenablefinal) seem to work correctly.
i was talking about http://sources.gentoo.org/viewcvs.py/gentoo-x86/app-office/lyx/lyx-1.6.0_beta3.ebuild and gcc 4.3. as far 1.5.6 unless you raise this issue on lyx devel list, its very unprobable this is going to be fixed.
> as far 1.5.6 unless you raise this issue on lyx devel list, its very unprobable > this is going to be fixed. > Could you please do it, cause you're already acknowledged there ?
i have already asked 4.3 people to confirm your last report and they haven't suceeded. if this doesn't reappear in 1.6 i'm not going to spend more time on it, so its up to you.
Well, only thing I've done was: EXTRA_ECONF="--without-included-boost --disable-pch" emerge lyx That was enough to trigger the bug. And as I said, it's `--disable-pch` part that's important, cause the precompiled header includes those headers correctly at global scope, while abort.cpp file includes them (it seems incorrectly) inside namespace lyx. That's why I had that problem in comment 4, cause there was no ::abort when those headers were included in lyx namespace.
pch is disabled in 1.6 ebuilds and we are not going to change it in 1.5 -> closing
(In reply to comment #22) > pch is disabled in 1.6 ebuilds and we are not going to change it in 1.5 -> > closing moreover i've had build problems when enabling pch in 1.6 series; but let's wait until 1.6 is officially out to close this one as for now the best visible lyx version still has the issue.
(In reply to comment #23) > (In reply to comment #22) > > pch is disabled in 1.6 ebuilds and we are not going to change it in 1.5 -> > > closing > > moreover i've had build problems when enabling pch in 1.6 series; but let's > wait until 1.6 is officially out to close this one as for now the best visible > lyx version still has the issue. > lyx 1.6.0 is out and unmasked -> fixed
