Re: Partial solution to slow null build of scons

2006-05-13 Thread Abdelrazak Younes
Abdelrazak Younes a écrit : Enrico Forestieri a écrit : On Fri, May 12, 2006 at 08:51:48PM +0200, Abdelrazak Younes wrote: It is very fast (faster than make install) but I think it is slowed down but the console display, I would suggest to only echo directory install, not every files (there a

Re: Partial solution to slow null build of scons

2006-05-13 Thread Jean-Marc Lasgouttes
> "make install" always install everything here, > I don't know if that's intended or if that's one of my perpetual > problems with autotools. This is intended. JMarc

Re: Partial solution to slow null build of scons

2006-05-13 Thread Abdelrazak Younes
Enrico Forestieri a écrit : On Fri, May 12, 2006 at 08:51:48PM +0200, Abdelrazak Younes wrote: It is very fast (faster than make install) but I think it is slowed down but the console display, I would suggest to only echo directory install, not every files (there are a lot of xpm and example f

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 08:51:48PM +0200, Abdelrazak Younes wrote: > It is very fast (faster than make install) but I think it is slowed down > but the console display, I would suggest to only echo directory install, > not every files (there are a lot of xpm and example files). Please don't do

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 01:22:00PM -0500, Bo Peng wrote: > > scons install tried to compile lyxclient which it should not try... > > because it fails. > > Is it destined to fail? Then, I need to handle this. It is simple. If the fcntl test fails then don't try to compile lyxclient. > Also, does

Re: Partial solution to slow null build of scons

2006-05-12 Thread Abdelrazak Younes
Bo Peng a écrit : Then try 'scons install' again. (I would like to see if Resources is corrected installed. Success except for two problems: 1) man is installed in Resouces instead of Resources (typo I presume) 2) locales are not installed (I think you know about that). It is very fast (faster

Re: Partial solution to slow null build of scons

2006-05-12 Thread Abdelrazak Younes
Bo Peng a écrit : scons install tried to compile lyxclient which it should not try... because it fails. Is it destined to fail? Then, I need to handle this. With autotools, it is destined to fail if the test of socket.h fail. Please remove client from env.Install(env['BIN_DIR'], [lyx, tex

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
scons install tried to compile lyxclient which it should not try... because it fails. Is it destined to fail? Then, I need to handle this. Please remove client from env.Install(env['BIN_DIR'], [lyx, tex2lyx, client]) and change if 'client' in targets or 'install' in targets: to if 'client

Re: Partial solution to slow null build of scons

2006-05-12 Thread Abdelrazak Younes
Bo Peng a écrit : It worked perfectly. I don't know why but the moced files were recompiled again and then the final link step succeeded without a glitch. I installed and tested it. Everything fine, lyxclient works, too. Then, there must be some scons files (or intermediate files) lying around

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 11:10:00AM -0500, Bo Peng wrote: > >It worked perfectly. I don't know why but the moced files were > >recompiled again and then the final link step succeeded without > >a glitch. I installed and tested it. Everything fine, lyxclient > >works, too. > > Then, there must be so

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
It worked perfectly. I don't know why but the moced files were recompiled again and then the final link step succeeded without a glitch. I installed and tested it. Everything fine, lyxclient works, too. Then, there must be some scons files (or intermediate files) lying around my cygwin tree. (Or

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 10:13:08AM -0500, Bo Peng wrote: > >Yesterday I succeeded, but I am not sure it was exactly before the > >latest patch. > > I guess the first thing to check is again if libqt3.a has the correct > number of entries. That is to say, if qt3/moced_files are actually > compiled

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
Yesterday I succeeded, but I am not sure it was exactly before the latest patch. I guess the first thing to check is again if libqt3.a has the correct number of entries. That is to say, if qt3/moced_files are actually compiled in. This manual mocing thing was always the source of trouble. You m

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 10:00:22AM -0500, Bo Peng wrote: > >> I just submitted another patch. Please 'svn update' and try again. I > >> force the use of $CXX on all platform, that should fix the solaris > >> problem. My cygwin complains during linking again, and I do not see > >> why. :-( > > > >I

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
> I just submitted another patch. Please 'svn update' and try again. I > force the use of $CXX on all platform, that should fix the solaris > problem. My cygwin complains during linking again, and I do not see > why. :-( I am going to test it. Do you succeed *before* the latest patch? I mean, i

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 09:26:49AM -0500, Bo Peng wrote: > I just submitted another patch. Please 'svn update' and try again. I > force the use of $CXX on all platform, that should fix the solaris > problem. My cygwin complains during linking again, and I do not see > why. :-( I am going to test

Re: Partial solution to slow null build of scons

2006-05-12 Thread Abdelrazak Younes
Bo Peng a écrit : Abdel, please test also. This is not yet weekend so I will be able to fix a few bugs. Did you fix my reported bug? Abdel.

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
That's the stuff in intl/ and po/ and this is what is to be done (unless --disable-nls is used): OK. Let us wait till the po stuff is added. > I only find lyx.man, what is the processing step? cp lyx.man $PREFIX/man/man1/lyx.1 cp development/lyxsocket/lyxclient.man $PREFIX/man/man1/lyxclient.1

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 08:33:55AM -0500, Bo Peng wrote: > >$PREFIX/bin (lyx, lyxclient, tex2lyx) > >$PREFIX/share/lyx(the lyx sysdir) > >$PREFIX/share/locale (all the locale stuff goes here) > > Where is it? That's the stuff in intl/ and po/ and this is what

Re: Partial solution to slow null build of scons

2006-05-12 Thread Abdelrazak Younes
As you seem to be active before the week-end, here is a bug report for you ;-) : scons CCFLAGS=-O3 qt_dir=d:/program/Qt/4.1 extra_inc_path=d:/program/Aspell-0.60.4/include extra_lib_path=d:/pr ogram/Aspell-0.60.4/lib logfile=logfile.log mode=release [...] scons: *** Source `D:\cygwin\home\yn

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
$PREFIX/bin (lyx, lyxclient, tex2lyx) $PREFIX/share/lyx(the lyx sysdir) $PREFIX/share/locale (all the locale stuff goes here) Where is it? $PREFIX/man/man1 (lyx.1, lyxclient.1, tex2lyx.1) I only find lyx.man, what is the processing step? Bo

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 08:11:51AM -0500, Bo Peng wrote: > >Same result, both on cygwin: > > So, solaris compile succeed this time right? I will ask for a better > solution on scons-user. I solved the problem passing CC=g++. Perhaps I should have used LINK=g++ but I don't know scons. Would that w

Re: Partial solution to slow null build of scons

2006-05-12 Thread Bo Peng
Same result, both on cygwin: So, solaris compile succeed this time right? I will ask for a better solution on scons-user. Using my previous patch I was able to install lyx. Everything went ok, only the man pages were not installed. I did not do it because lyx can run without them. It should

Re: Partial solution to slow null build of scons

2006-05-12 Thread Enrico Forestieri
On Fri, May 12, 2006 at 12:02:44AM -0500, Bo Peng wrote: > > > >I introduced an exe_suffix. > > This is the last thing you want to do with a scons system. The copy > business is surprisingly tricky, and I get the code from the user's > guide. Please have a look at the attached patch. > > >> >On s

Re: Partial solution to slow null build of scons

2006-05-12 Thread Abdelrazak Younes
Bo Peng a écrit : Now, unless I get some bug reports, I will assume that scons is basically working and I will take this weekend off. It's been a bit intense for me lately and I own my kids a few hours of playing time. They deserves their daddy! The bug reports could wait until next week, ther

Re: Partial solution to slow null build of scons

2006-05-11 Thread Bo Peng
I introduced an exe_suffix. This is the last thing you want to do with a scons system. The copy business is surprisingly tricky, and I get the code from the user's guide. Please have a look at the attached patch. > >On solaris scons uses gcc for linking, so there are tons of undefined > >refe

Re: Partial solution to slow null build of scons

2006-05-11 Thread Enrico Forestieri
On Thu, May 11, 2006 at 06:30:49PM -0500, Bo Peng wrote: > >Please, find attached a small bug fix for lyxclient. > > So socket_libs should be separated from system_libs anyway? You know, > for lyxclient, system_libs has more stuff than socket_libs... > > Please, if you think socket_libs should be

Re: Partial solution to slow null build of scons

2006-05-11 Thread Bo Peng
BTW, you might be interested in adding a no-config=1 or no-detect=1 option to bypass the generation of config.h, or the whole detection process. (keep only those that will have a *useful* value in env.) Generating config.h each time is kind of silly, even if the process is relatively quick with c

Re: Partial solution to slow null build of scons

2006-05-11 Thread Bo Peng
Another small fix to avoid problems with library ordering: Your patch looks fine. Thanks. Bo

Re: Partial solution to slow null build of scons

2006-05-11 Thread Bo Peng
Please, find attached a small bug fix for lyxclient. So socket_libs should be separated from system_libs anyway? You know, for lyxclient, system_libs has more stuff than socket_libs... Please, if you think socket_libs should be separated, provide an updated patch that 1. lyxclient only uses so

Re: Partial solution to slow null build of scons

2006-05-11 Thread Enrico Forestieri
On Fri, May 12, 2006 at 12:56:44AM +0200, Enrico Forestieri wrote: > Please, find attached a small bug fix for lyxclient. Another small fix to avoid problems with library ordering: g++ -o release/qt3/lyx -Lrelease/libs -L/opt/qt/lib -L/opt/qt/lib -L/opt/lib -L/usr/X11/lib -L. -llyxbase_pre -lma

Re: Partial solution to slow null build of scons

2006-05-11 Thread Enrico Forestieri
On Thu, May 11, 2006 at 04:30:33PM -0500, Bo Peng wrote: > $ scons lyxbase > $ scons [prefix=/path/to/lyx] [exec_path=/path/to/lyx_bin] install > are in. > > So my lastest fresh run (linux) is like > $ scons client > $ scons tex2lyx > $ scons lyxbase > $ scons qt3 > $ scons frontend=qt4 qt_dir=/p

Re: Partial solution to slow null build of scons

2006-05-11 Thread Bo Peng
$ scons tex2lyx (or boost mathed frontends controllers qt3 qt4 client graphics) They correspond to individual libraries (executable) and because I manage to avoid *processing* other directories, the null-build time for them are usually 2 to 3 seconds. The qt3/qt4 part, does take around 15s becaus

Partial solution to slow null build of scons

2006-05-11 Thread Bo Peng
Dear all, I have done a thorough profiling on the starting process of scons, and it turns out that there is no way (as of now) to reduce the null build time further, except for a few command line options that can reduce the time from 32 to 26s (assuming no dependence change etc). I however, can