LDFLAG --as-needed / Speeding up of GNOME or KDE

Message

taipan67 · Post by **taipan67** » Sun Jul 03, 2005 6:55 pm

Rainmaker wrote:
Code: Select all
ld: cannot find -lplc4

I think that means that the linker, 'ld', can't find the 'plc4' library to link against. You'll see confusing stuff like that every time you emerge something; Stuff prefixed with a '-l' (lowercase L) refers to a library to be linked against, stuff prefixed with a '-I' (uppercase i) refers to a header-file to be included.

On my system, i've got two 'libplc4.so' shared library files; One in each of /usr/lib/MozillaFirefox & /usr/lib/MozillaThunderbird... No idea what 'gnome-panel' needs them for, though.

As for 'udev', my entire system has been built from scratch with this FLAG (i just comment it out of /etc/make.conf if i get an individual failure), & i've had no problems like yours. Maybe it's something unrelated... Sorry i don't have any more ideas, just now.

Rainmaker · Post by **Rainmaker** » Sun Jul 03, 2005 7:03 pm

that's probably it:

Code: Select all

Medusa# locate plc4
/home/dannie/sunbird/libplc4.so
/opt/mozilla/libplc4.so
/opt/sunbird/libplc4.so
/usr/lib/MozillaFirefox/libplc4.so
/usr/lib/MozillaThunderbird/libplc4.so
/usr/lib/nspr/libplc4.so
/usr/lib/nspr/libplc4.a
/usr/lib/nvu/libplc4.so
/usr/share/gecko-sdk/lib/libplc4.so

So it's not related to this flag, but proabably a missing include diror something. I'll expreriment some more. Thanks for the tip.

taipan67 · Post by **taipan67** » Sun Jul 03, 2005 7:28 pm

Rainmaker wrote:...So it's not related to this flag, but proabably a missing include diror something.

Sorry - as usual, i haven't explained myself as clearly as i'd like...

I'd imagine that the failure of 'gnome-panel' is directly attributable to this FLAG, as it affects the linker itself, & seems to have different ideas from the source-code about what is & isn't a dependant shared-library. If it was me, i'd re-emerge it without 'Wl,--as-needed', if only to see if it will...

When i said that i didn't think the FLAG was to blame, i was referring to your problem with 'udev' only - sorry again for the misunderstanding.

Rainmaker · Post by **Rainmaker** » Sun Jul 03, 2005 8:12 pm

Rainmaker wrote:Removing the flag does NOT solve it this time. I'm having the same problem with bind-tools.

you were clear enough, I got it

The problem was I emerged mozilla-bin instead of mozilla and put mozilla in my portage/package.provided.

Unemerging -bin and emergeing mozilla fixed it

Thanks for the hint.It helped me figure out where the problem was.

The problem with udev was solved for me when emerging it without --as-needed. Just wanted to note that if others were having similair problems.

KD-120RD · Post by **KD-120RD** » Sun Jul 03, 2005 9:10 pm

Hi

I recopiled my kde with --ass-needed without errors. Kde-3.4.1 (without any session apps) starts about 2 secs faster. I also made some benchmarks for kde apps startuptimes, with "$ time [APP] &". Each time was taken 3 times. The resultes are average values.

Befor:
APP (size) | First start | Sec. start
kmail (15472b) | 3.71 sec | 1.9 sec
kopete (679kb) | 2.1 sec | 1.06 sec
kcontrol (4,7kb) | 1.28 sec | 0.97 sec

After:
APP (size) | First start | Sec. start
kmail (13200b) | 3.23 sec | 1.35 sec
kopete (678kb) | 2.05 sec | 0.9 sec
kcontrol (3,7kb) | 1.23 sec | 0.86 sec

I would say that 0.5 sec can be felt in normal pc usage. So generally its a good idea to use --ass-needed with kde.
Now I want to try it on some nonkde apps, like firefox or even glibc. But can someone post how exactly ldflags="" must look like to use --enable-new_ldflags? ./configure fails with

Code: Select all

checking for C compiler default output file name... configure: error: C compiler cannot create executables

when it looks like one of these

Code: Select all

LDFLAGS="-Wl,-O1 -Wl,--sort-common -s --enable-new_ldflags"
LDFLAGS="-Wl,-O1 -Wl,--sort-common -s Wl,--enable-new_ldflags"

Adrian

rhill · Post by **rhill** » Mon Jul 04, 2005 1:28 am

devsk wrote:one more thing is configure arg is --enable-new_ldflags and NOT --enable-new-ldflags as mentioned by the ./configure --help. bollucks knew about it but never shared it...that's a shame.

if it is enable-new_ldflags then it's a typo. try a google search for both.

neuron · Post by **neuron** » Mon Jul 04, 2005 12:34 pm

anyone know how many packages support --enable-new_ldflags?

I was thinking about rebuilding gnome with as-needed, but it'd be much nicer and safer to use --enable-new_ldflags wouldn't it?

infirit · Post by **infirit** » Mon Jul 04, 2005 2:11 pm

neuron wrote:anyone know how many packages support --enable-new_ldflags?

I was thinking about rebuilding gnome with as-needed, but it'd be much nicer and safer to use --enable-new_ldflags wouldn't it?

I always use --as-needed by default and only disable it for the packages that do not compile or are know to not work, original post. If you find new packages please post them and I will update it.

Most KDE packages us the the safe --enable-new_ldflags but I do not think many others do...

taipan67 · Post by **taipan67** » Mon Jul 04, 2005 2:44 pm

infirit wrote:I always use --as-needed by default and only disable it for the packages that do not compile or are know to not work...

I've been doing the same, but am thinking that for my next system build, i'm gonna reverse that policy, & just enable it to rebuild the main app's, after checking file-sizes & start-times for comparison.

This is as much for peace of mind as anything else... I've had a few dependencies that compiled successfully, but when i've checked the logs, they've found, like 'imlib', that they couldn't check for availability of '>=gtk-1.2.*', for example, & carried on their compilation by not including the features that depend on it - 'media-libs/smpeg-0.4.4-r5' is a case in point. I only discovered the discrepancy because i was emerging 'one-package-at-a-time' for fear of missing just such an occurence, & for meta-packages that's not very practical.

I've also become slightly paranoid that my use of this flag globally might be responsible for 'mail-client/mozilla-thunderbird-1.0.2-r1' incorrectly believing that it was compiled without the 'crypt' flag enabled, hence no 'enigmail', & possibly why my mail-filters aren't working properly...

As i said, it's probably overly paranoid, but i'll only be sure by re-compiling from scratch, so it'll be a week or so before that happens on my other partition...

KD-120RD · Post by **KD-120RD** » Mon Jul 04, 2005 4:57 pm

Hi,

if it is enable-new_ldflags then it's a typo. try a google search for both.

Did that. Result: enable-new_ldflags 3 hits, enable-new-ldflags 87 hits and found things like this:

>--enable-new-ldtags or --enable-new-ldflags? I use the latter since
> quite a while, so if that works I'd be very happy

The latter.

--enable-new-ldflags in configure turns on --enable-new-dtags and
--as-needed for ld

So I would say --enable-new-ldflags is the right one. (Will probably be killed for that one by infirit

) Anyhow none of them work! I wonder if anyone got it to work at all...

rm · Post by rm » Tue Jul 05, 2005 6:39 pm

=gnome-base/orbit-0.5.17 does not compile with the --as-needed setting.
Without, it just compiles fine.

roel

emerge --info gives

Code: Select all

Portage 2.0.51.19 (default-linux/x86/2005.0, gcc-4.0.1-pre20050702, glibc-2.3.5.20050421-r0, 2.6.11-love2 i686)
=================================================================
System uname: 2.6.11-love2 i686 AMD Athlon(tm) XP 2500+
Gentoo Base System version 1.6.12
Python:              dev-lang/python-2.4.1-r1 [2.4.1 (#1, Jul  2 2005, 00:03:40)]
dev-lang/python:     2.4.1-r1
sys-apps/sandbox:    [Not Present]
sys-devel/autoconf:  2.13, 2.59-r6
sys-devel/automake:  1.5, 1.8.5-r3, 1.6.3, 1.7.9-r1, 1.4_p6, 1.9.5
sys-devel/binutils:  2.16.91.0.1
sys-devel/libtool:   1.5.16
virtual/os-headers:  2.6.11-r2
ACCEPT_KEYWORDS="x86"
AUTOCLEAN="yes"
CFLAGS="-Os -march=athlon-xp -m3dnow -msse -fomit-frame-pointer -pipe -fforce-addr -momit-leaf-frame-pointer -fno-ident -ftracer -funit-at-a-time"
CHOST="i686-pc-linux-gnu"
CONFIG_PROTECT="/etc /usr/kde/2/share/config /usr/kde/3/share/config /usr/lib/X11/xkb /usr/share/config /var/qmail/control"
CONFIG_PROTECT_MASK="/etc/gconf /etc/terminfo /etc/env.d"
CXXFLAGS="-Os -march=athlon-xp -m3dnow -msse -fomit-frame-pointer -pipe -fforce-addr -momit-leaf-frame-pointer -fno-ident -ftracer -funit-at-a-time -fvisibility-inlines-hidden"
DISTDIR="/usr/portage/distfiles"
FEATURES="autoaddcvs autoconfig ccache distlocks sandbox sfperms strict"
GENTOO_MIRRORS="ftp://ftp.belnet.be/linux/gentoo/"
LANG="en_US.UTF-8"
LC_ALL="nl_BE.UTF-8"
LDFLAGS="-Wl,-O2 -Wl,--as-needed -Wl,--sort-common"
MAKEOPTS="-j1"
PKGDIR="/usr/portage/packages"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY="/usr/local/portage"
SYNC="rsync://rsync.gentoo.org/gentoo-portage"
USE="x86 3dnow X acpi alsa apache2 apm arts avi berkdb bitmap-fonts cairo cdr crypt cups dvd dvdr emboss encode esd flac font-server foomaticdb fortran gdbm gif gnome gpm gtk gtk2 gtkhtml hal imlib innodb ipv6 ithreads jpeg libg++ libwww lm_sensors mad mikmod mmx mono motif mozdevelop mp3 mpeg mpm-worker ncurses nls nptl nptlonly objc odbc ofx ogg oggvorbis openal opengl oss pam pdflib perl png posix pthreads python qt quicktime readline scanner sdl spell sse ssl svg tcpd threads tiff truetype truetype-fonts type1-fonts unicode usb userlocales vhosts win32codecs wmf xinerama xml2 xmlrpc xmms xpm xprint xv zlib userland_GNU kernel_linux elibc_glibc"
Unset:  ASFLAGS, CBUILD, CTARGET, LINGUAS

taipan67 · Post by **taipan67** » Tue Jul 05, 2005 7:00 pm

This is slightly Off-Topic, but it seems likely to have a captive audience, so...

I've noticed that quite a number of people are using values greater than one with the 'Wl,-O*' LDFLAG. My understanding of the 'ld man-page' has always been that this particular FLAG is as good as boolean, having only two states:- zero & not-zero.

Can anybody provide any clarification of this, please? And also, have users of values greater than one experienced any noticeable gains from doing so?

neuron · Post by **neuron** » Tue Jul 05, 2005 7:01 pm

taipan67 wrote:This is slightly Off-Topic, but it seems likely to have a captive audience, so...

I've noticed that quite a number of people are using values greater than one with the 'Wl,-O*' LDFLAG. My understanding of the 'ld man-page' has always been that this particular FLAG is as good as boolean, having only two states:- zero & not-zero.

Can anybody provide any clarification of this, please? And also, have users of values greater than one experienced any noticeable gains from doing so?

your entirly right and they are wrong

, there's no reason to use O2 here whatsoever, it does the same.

infirit · Post by **infirit** » Tue Jul 05, 2005 7:27 pm

neuron wrote:your entirly right and they are wrong , there's no reason to use O2 here whatsoever, it does the same.

If I'm using -O9 am wrong or am I correct differently

infirit · Post by **infirit** » Tue Jul 05, 2005 8:17 pm

KD-120RD wrote:Hi,
if it is enable-new_ldflags then it's a typo. try a google search for both.
Did that. Result: enable-new_ldflags 3 hits, enable-new-ldflags 87 hits and found things like this:

>--enable-new-ldtags or --enable-new-ldflags? I use the latter since
> quite a while, so if that works I'd be very happy

The latter.

--enable-new-ldflags in configure turns on --enable-new-dtags and
--as-needed for ld
So I would say --enable-new-ldflags is the right one. (Will probably be killed for that one by infirit ) Anyhow none of them work! I wonder if anyone got it to work at all...

Nah, there enough killing in the world already

Actually I do not use --enable-new-/_ldflags but have been experimenting on konsole. With -, _ or without all have the same results. Also --as-needed did nothing

I'l do some more testing and try to figure this out.

taipan67 · Post by **taipan67** » Tue Jul 05, 2005 8:36 pm

infirit wrote:
neuron wrote:your entirly right and they are wrong , there's no reason to use O2 here whatsoever, it does the same.
If I'm using -O9 am wrong or am I correct differently

Let's see what 'man ld' has to say on the topic...

-O level
If level is a numeric values greater than zero ld optimizes the output. This might take significantly longer and therefore probably should only be enabled for the final binary.

Well what fsckin' use is that

Now do you see why i asked in the first place? This could get ugly...

infirit · Post by **infirit** » Tue Jul 05, 2005 9:06 pm

infirit wrote:
KD-120RD wrote:Hi,

So I would say --enable-new-ldflags is the right one. (Will probably be killed for that one by infirit ) Anyhow none of them work! I wonder if anyone got it to work at all...
Nah, there enough killing in the world already

Actually I do not use --enable-new-/_ldflags but have been experimenting on konsole. With -, _ or without all have the same results. Also --as-needed did nothing

I'l do some more testing and try to figure this out.

I made a typo when testing konsole with --as-needed before now it works again, 2 linked instead of 40

--enable-new_ldflags and --enable-new-ldflags do not activate --as-needed for konsole

I also tried --enable-new-dtags/_dtags, no --as-needed.

So if anyone figures this out, let us know...

KD-120RD · Post by **KD-120RD** » Tue Jul 05, 2005 9:53 pm

Thx for not killing me, infirit.

Afaik, --enable-new-dtags/_dtags does not enable --as-needed. It is another optimising option. But I'm not 100% positiv. Will check this the next days. Will also do some research on --enable-new-ldflags/_ldflags then.

Will go zzZZzz now, tomorrow exam ...

wel · Post by **wel** » Fri Jul 08, 2005 8:39 am

X.org 6.8.1-r2 compiled here with --as-needed. Two days after, it's rock solid.

Rainmaker · Post by **Rainmaker** » Fri Jul 08, 2005 11:55 pm

I'm having trouble emerging the latest k3b version (0.12.2). Not sure if it's a gcc4 or a --as-needed issue, but the ./configure dies with this error:

checking if UIC has KDE plugins available... no
configure: error:
you need to install kdelibs first.

If you did install kdelibs, then the Qt version that is picked up by
this configure is not the same version you used to compile kdelibs.
The Qt Plugin installed by kdelibs is *ONLY* loadable if its the
same Qt version, compiled with the same compiler and the same Qt
configuration settings.

I've recompiled kdelibs, qt-3.3.4 and qt-4.0.0 enlesly now, but I'm not sure where to start looking... If anyone has a clue, please let me know.

*edit: fixed.

This was an issue with GCC 4.1.0 beta.

emerge -e system with gcc 3.3.4, then I emerged gcc-4.0.1
I recompiled system again with 4.0.1, then re-emerged kdeibs and qt-3.3.4-r6. After this, k3b worked again

darkless · Post by **darkless** » Sun Jul 10, 2005 1:06 pm

taipan67 wrote:I've also become slightly paranoid that my use of this flag globally might be responsible for 'mail-client/mozilla-thunderbird-1.0.2-r1' incorrectly believing that it was compiled without the 'crypt' flag enabled, hence no 'enigmail', & possibly why my mail-filters aren't working properly...

As i said, it's probably overly paranoid, but i'll only be sure by re-compiling from scratch, so it'll be a week or so before that happens on my other partition...

I've seen the same behavior from gaim when using the flag "gnutls". I couldn't figure out why it wouldn't allow me to use the MSN protocol. A 'gaim -d' later, I saw that SSL support hadn't been compiled in, and watching a reemerge confirmed that it didn't detect the gnutls libraries. Emerging gaim without the --as-needed flag made it work again

taipan67 · Post by **taipan67** » Sun Jul 10, 2005 2:18 pm

darkless wrote:
taipan67 wrote:I've also become slightly paranoid that my use of this flag globally might be responsible for 'mail-client/mozilla-thunderbird-1.0.2-r1' incorrectly believing that it was compiled without the 'crypt' flag enabled, hence no 'enigmail', & possibly why my mail-filters aren't working properly...

As i said, it's probably overly paranoid, but i'll only be sure by re-compiling from scratch, so it'll be a week or so before that happens on my other partition...
I've seen the same behavior from gaim when using the flag "gnutls". I couldn't figure out why it wouldn't allow me to use the MSN protocol. A 'gaim -d' later, I saw that SSL support hadn't been compiled in, and watching a reemerge confirmed that it didn't detect the gnutls libraries. Emerging gaim without the --as-needed flag made it work again

Thanks for the follow-up, darkless - although 'net-im/gaim' has been added to the list of packages that fail in the 1st post on this thread, it might be worth listing the full ebuild-title (with numbers), just in case it's a problem unique to your release (which doesn't sound likely).

On the 'mail-client/mozilla-thunderbird-1.0.2-r1' front, the mail-filters weren't working because i'd configured them wrong.

...And regarding 'enigmail', being inspired to dig a bit deeper, i've just run this test :-

Code: Select all

gentoo taipan # cd /var/log/portage
gentoo portage # ls -alFh *thunderbird*
-rw-r--r--  1 root root 3.9M Jun 30 02:08 2657-mozilla-thunderbird-1.0.2-r1.log
-rw-r--r--  1 root root  337 Jun 30 02:08 2658-mozilla-thunderbird-1.0.2-r1.log
gentoo portage # less 2657-*thunderbird* | grep -i enigmail
......

...and discovered amongst a few hundred lines of output that 'enigmail' was built as requested, but none of the relevant 'chrome' components could be found - so it's there, it just can't be accessed from the GUI.

As of this moment, i don't know if that's because of the FLAG, or because of a bug in the source-tarball, & to find out i'll have to either figure out how to safeguard my existing email & account-settings, then re-compile without the FLAG, or burn shit-loads of personal files from my other installation to disk & build it without the FLAG on that side of the hard-drive.

I'm leaning towards the latter, coz i need to get round to that anyway, but i'll be sure to post my findings either way...

aminalshmu · Post by **aminalshmu** » Sun Jul 10, 2005 7:16 pm

media-libs/gle-3.0.1-r2 fails with --as-needed, but works fine without.

Code: Select all

jordan@jordan jordan $ emerge info
Portage 2.0.51.22-r1 (default-linux/x86/2005.0, gcc-4.0.1, glibc-2.3.5.20050421-r0, 2.6.12-morph5 i686)
=================================================================
System uname: 2.6.12-morph5 i686 Intel(R) Pentium(R) 4 CPU 2.80GHz
Gentoo Base System version 1.6.12
distcc 2.18.3 i686-pc-linux-gnu (protocols 1 and 2) (default port 3632) [disabled]
ccache version 2.4 [enabled]
dev-lang/python:     2.3.5, 2.4.1-r1
sys-apps/sandbox:    1.2.10
sys-devel/autoconf:  2.13, 2.59-r7
sys-devel/automake:  1.4_p6, 1.5, 1.6.3, 1.7.9-r1, 1.8.5-r3, 1.9.5
sys-devel/binutils:  2.16.1
sys-devel/libtool:   1.5.18-r1
virtual/os-headers:  2.6.11-r2
ACCEPT_KEYWORDS="x86 ~x86"
AUTOCLEAN="yes"
CBUILD="i686-pc-linux-gnu"
CFLAGS="-O2 -march=pentium4 -fomit-frame-pointer -pipe"
CXXFLAGS="-O2 -march=pentium4 -fomit-frame-pointer -pipe"
CHOST="i686-pc-linux-gnu"
LDFLAGS="-Wl,-O1 -Wl,--as-needed"

robnotts · Post by **robnotts** » Tue Jul 12, 2005 6:44 pm

Can I add mail-client/evolution-2.2.1.1 to the list of applications that fails with --as-needed. It fails saying the pilot-link needs compiling with UTF-8 conversion support. Compiles fine without the --as-needed flag.

Rob.

PS: Pretty impressed with the changes made to the perceived speed of the computer when kde (all of it) is compiled with the --as-needed flag.

rm · Post by rm » Sat Jul 16, 2005 6:58 pm

net-www/apache-2.0.54-r12 builds with --as-needed,
but installing mod_python-3.1.4-r1 fails

rebuilding apache without --as-needed, and then installing mod_python works.

bye,
roel

ps: didn't try to start apache before rebuilding, sorry, this is only an afterthought