From: Rafael L. <rla...@us...> - 2004-01-05 19:24:59
|
* Alan W. Irwin <ai...@us...> [2004-01-05 07:52]: > In sum it is an on-going difficult situation where the rules may change > from time to time on various platforms, but we just have to keep track of > the situation and only change over to AC_F77_LIBRARY_LDFLAGS for a > particular platform when we are forced to do so. At the present time the > score is zero platforms where it is necessary/useful. Okay, I went ahead and commented the call to AC_F77_LIBRARY_LDFLAGS in configure.ac. I put a comment in the file explaining why. -- Rafael |
From: Koen v. d. D. <kvd...@ea...> - 2004-01-08 03:32:59
|
On Jan 5, 2004, at 10:52 AM, Alan W. Irwin wrote: > Having slept on the situation I think we can take an even more > conservative > approach. We have no evidence now that AC_F77_LIBRARY_LDFLAGS does > anything > positive on any platform (despite the documentation saying different). > It > I have done some more testing on this matter. If AC_F77_LIBRARY_LDFLAGS is set then on my machine then FLIBS is set to '-lcrt.o -lfrtbegin -lSystem - lg2c.' If I set FLIBS manually (using export) without '-lg2c', compilation goes fine. Apparently lg2c causes the compilation to stop. This is all on Mac OS X and the 20040104 tarball. - Koen. |
From: Rafael L. <rla...@us...> - 2004-01-05 12:23:29
|
* Michel Peyrard <Mic...@en...> [2004-01-05 13:16]: > However I noticed that, when I run x01f with X Window device, there is no > way to exit from the program: pressing Return on the plot window has no > effect (while it used to close the window in earlier versions of plplot) > and CTRL C in the window from which I launched the program has no effect > either. The only possibility to stop is to close the launching window. Does this happen also with x01c? -- Rafael |
From: Michel P. <Mic...@en...> - 2004-01-05 12:28:47
|
> > > However I noticed that, when I run x01f with X Window device, there is no > > way to exit from the program: pressing Return on the plot window has no > > effect (while it used to close the window in earlier versions of plplot) > > and CTRL C in the window from which I launched the program has no effect > > either. The only possibility to stop is to close the launching window. > > Does this happen also with x01c? > I had not tried, but I checked: the answer is YES. Michel |
From: Rafael L. <rla...@us...> - 2004-01-05 15:28:04
|
* Alan W. Irwin <ai...@us...> [2004-01-04 16:40]: > Rafael, do you plan an official release candidate made as a file release on > SF? That gets substantial additional publicity beyond what you get from > announcing it on list, and it also focusses the developers attention on > testing as well. If you do decide to do this, I would suggest doing it in > the 3rd week of January about a week before your target date for the real > release. Last time we did an official release candidate we got something > like 100 downloads in that "test week" which I think is more than enough > testers (presumably most of them new) to make it worthwhile. Note, none of > those testers reported any problems, but on the other hand, I was happy we > had that extra insurance that bugs in 5.2.1 (beyond the known issues) were > hard to find. I am not sure I will do release candidate as a file release on SF. I have the feeling that it is too much hassle for few benefits, since people tracking the SF announcement tend to be rather "users" than "testers". (unlike people hanging on plplot-devel). Free time will be scarce for me during the next weeks. Besides, I always do a simultaneous release of the Debian packages when a cvs tarball comes out (I sort of automated the process). It increases the chances of the release being tested, since Debian people are notoriously "picky" :-) -- Rafael |
From: Koen v. d. D. <kvd...@ea...> - 2004-01-06 00:53:27
|
On Jan 5, 2004, at 12:38 AM, Alan W. Irwin wrote: > fink g77 3.4 is somewhat problematic at this time, but I think Michel > found > some workarounds to make it work before so I think those will be > necessary > until the fink g77 3.4 problems are resolved. For me it compiled without problems with the 20031231 tarball, Michel still had problems, IIRC. > Note this time we have > reintroduced AC_F77_LIBRARY_LDFLAGS which adds additional link flags > that we > are told by the documentation are supposed to facilitate allowing most > Unix > platforms to link the mixed C and fortran library libplplotf77 Is there a way I can try that out before a new tarball comes out? I commented out the line AC_F77_LIBRARY_LDFLAGS and according to one of your earlier mails, then ran ./bootstrap.sh, followed by ./configure, etc. But I got an error about 'missing --run autoheader'. Is there something else I need to do before the usual ./configure, etc? I am also going to see how other fink packages that use g77 are packaged, maybe that can be of any help. thanks, - Koen. |
From: Alan W. I. <ai...@us...> - 2004-01-06 03:12:34
|
On 2004-01-05 19:52-0500 Koen van der Drift wrote: > Is there a way I can try that out before a new tarball comes out? Yes. In the case when you modify configure.ac (or sysloc.in or any of the Makefile.am files) in an unpacked cvs snapshot tarball you must do the following: * Make sure you have access to autoconf-2.58, automake-1.7.9, and libtool-1.5 packages. The exact versions are important for PLplot configuration to work, and I doubt the fink versions will do the job. You can download tarball packages for all of these from http://ftp.gnu.org/gnu/autoconf/, http://ftp.gnu.org/automake, and http://ftp.gnu.org/gnu/libtool. unpack, configure, make, and make install them all with the same special prefix (I chose /home/software/autotools/install), and make sure that /home/software/autotools/install/bin (or whatever prefix you choose) is first on your path before you start e.g., "which autoconf" should give you the tarball version not your system version. * In the freshly unpacked plplot directory execute ./bootstrap.sh The output of that command should be similar to the following ./bootstrap.sh Using aclocal options: -I /home/software/autotools/install/share/libtool/libltdl Running aclocal (GNU automake) 1.7.9... done Running autoheader (GNU Autoconf) 2.58... done Running libtoolize (GNU libtool) 1.5... done Running automake (GNU automake) 1.7.9... done Running autoconf (GNU Autoconf) 2.58... done Regenerating libltdl/aclocal+configure... done Note if that first line doesn't point to your special prefix area for the newly installed autotools you are cooked, and nothing will work. However, if that first line is right, then you can go ahead and ./configure; make; and make install as normal. Note the above instructions are almost identical to what you do with a fresh cvs checkout. However, the tarball version has some stuff prebuilt for your convenience. Thus, if you are using raw cvs instead you will need in addition access to swig-1.3.17 (exact version please from http://sourceforge.net/project/showfiles.php?group_id=1645) to help "make" automatically build key parts of the python and java interfaces. cvs access to PLplot can be obtained following the instructions at http://sourceforge.net/cvs/?group_id=2915. However, the anon read-only version you have access to sometimes runs up to a day late. Basically you have to check http://cvs.sourceforge.net/viewcvs.py/plplot/plplot/ to see what has made it to the anon cvs server. For example, if you look there now you will see Rafael's configure.ac change from 7 hours ago made it and is accessible to you and similarly for my recent post-tarball changes to examples/f77. Also, to get automatically informed of every cvs change you should join the plplot-cvs list. In sum, getting the right autotools and swig versions installed on your system gives you a chance to try all sorts of interesting things out and get timely access to all our cvs changes as well when it has been more than 24 hours since Rafael made a tarball. Alan __________________________ Alan W. Irwin email: ir...@be... phone: 250-727-2902 Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the PLplot scientific plotting software package (plplot.org), the Yorick front-end to PLplot (yplot.sf.net), the Loads of Linux Links project (loll.sf.net), and the Linux Brochure Project (lbproject.sf.net). __________________________ Linux-powered Science __________________________ |
From: <jc...@fe...> - 2004-01-05 20:00:13
|
On Monday 05 January 2004 06:19, Alan W. Irwin wrote: | On 2004-01-04 21:38-0800 Alan W. Irwin wrote: | > Because of the troubles we have encountered, we should also not | > just rely on the documentation that AC_F77_LIBRARY_LDFLAGS is | > supposed to work. In addition we should test that it works on as | > many Unix fortran platforms as possible. The SF compile farm is | > supposed to have a sparc8 processor available so I may try that in | > the next few days if they have the fortran compiler installed. | > Joao, will you try fortran on your OSF1 platform using the latest | > tarball? Yes. A quick test. But OSF1 is not a good target, as native fortran/c compiled/run OK with the previous setup (5.2.1). The configure cmd: F77=f77 ./configure --disable-static Some errors when doing make (in the examples/f build directory): bash-2.01$ for i in x??f.f; do f=`basename $i .f`; make $f; done f77 -g -c -o x01f.o `test -f 'x01f.f' || echo './'`x01f.f f90: Error: x01f.f, line 16: This name does not have a type, and must have an explicit type. [LNBLNK] & version(:lnblnk(version)) -----------------^ make: *** [x01f.o] Error 1 ... f77 -g -o .libs/x09f x09f.o ../../bindings/f77/.libs/libplplotf77d.so /usr/users1/deec/jcard/plplot-5.2.1.cvs.20040104/src/.libs/libplplotd.so /usr/users1/deec/jcard/plplot-5.2.1.cvs.20040104/lib/csa/.libs/libcsirocsa.so -lUfor -lfor -lFutil -lm -lots -lm_c32 ld: Unresolved: pl_setcontlabelparam_ f77: Severe: Failed while trying to link. make: *** [x09f] Error 1 ... f77 -g -c -o x17f.o `test -f 'x17f.f' || echo './'`x17f.f f90: Error: x17f.f, line 128: This name does not have a type, and must have an explicit type. [RAND] noise = rand(1) --------------^ make: *** [x17f.o] Error 1 And then all made demos runs OK under xwin, except that I had to CTRL-C them, as the <enter> key or <B2> mouse didn't finish the program. This is strange, as these keys/buttons does advance the plots, as in e.g. x04f. Ah! it also happens with the C demos (at least x01c) -- and also on Mac OSX., if I remember correctly. Could it be a pthreads issue? Yes, if I do ./x01c -dev xwin -drvopt usepth=0 ( a new driver option) the app ends as usual. But pthreads funtionality is working OK. More to debug. If I don't get the time, we have to ship with pthreads disabled. Joao | So we will have to rely mostly on Joao and anyone else with access to | Unix fortran to test the current cvs snapshot to give us some | additional information to help us decide whether to include | AC_F77_LIBRARY_LDFLAGS or comment it out for now. | | Alan | __________________________ | Alan W. Irwin | email: ir...@be... | phone: 250-727-2902 | | Astronomical research affiliation with Department of Physics and | Astronomy, University of Victoria (astrowww.phys.uvic.ca). | | Programming affiliations with the PLplot scientific plotting software | package (plplot.org), the Yorick front-end to PLplot (yplot.sf.net), | the Loads of Linux Links project (loll.sf.net), and the Linux | Brochure Project (lbproject.sf.net). | __________________________ | | Linux-powered Science | __________________________ | | | ------------------------------------------------------- | This SF.net email is sponsored by: IBM Linux Tutorials. | Become an expert in LINUX or just sharpen your skills. Sign up for | IBM's Free Linux Tutorials. Learn everything from the bash shell to | sys admin. Click now! | http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click | _______________________________________________ | Plplot-devel mailing list | Plp...@li... | https://lists.sourceforge.net/lists/listinfo/plplot-devel |
From: Alan W. I. <ai...@us...> - 2004-01-05 20:47:44
|
On 2004-01-05 19:58-0000 Jo=E3o Cardoso wrote: > On Monday 05 January 2004 06:19, Alan W. Irwin wrote: > | On 2004-01-04 21:38-0800 Alan W. Irwin wrote: > | > Because of the troubles we have encountered, we should also not > | > just rely on the documentation that AC_F77_LIBRARY_LDFLAGS is > | > supposed to work. In addition we should test that it works on as > | > many Unix fortran platforms as possible. The SF compile farm is > | > supposed to have a sparc8 processor available so I may try that in > | > the next few days if they have the fortran compiler installed. > | > Joao, will you try fortran on your OSF1 platform using the latest > | > tarball? > > Yes. A quick test. > But OSF1 is not a good target, as native fortran/c compiled/run OK with > the previous setup (5.2.1). The configure cmd: > > F77=3Df77 ./configure --disable-static > > Some errors when doing make (in the examples/f build directory): > > bash-2.01$ for i in x??f.f; do f=3D`basename $i .f`; make $f; done > > f77 -g -c -o x01f.o `test -f 'x01f.f' || echo './'`x01f.f > f90: Error: x01f.f, line 16: This name does not have a type, and must > have an explicit type. [LNBLNK] > & version(:lnblnk(version)) > -----------------^ > make: *** [x01f.o] Error 1 > > ... > > f77 -g -o .libs/x09f x09f.o ../../bindings/f77/.libs/libplplotf77d.so > /usr/users1/deec/jcard/plplot-5.2.1.cvs.20040104/src/.libs/libplplotd.so > /usr/users1/deec/jcard/plplot-5.2.1.cvs.20040104/lib/csa/.libs/libcsirocs= a.so > -lUfor -lfor -lFutil -lm -lots -lm_c32 > ld: > Unresolved: > pl_setcontlabelparam_ > f77: Severe: Failed while trying to link. > make: *** [x09f] Error 1 > > ... > > f77 -g -c -o x17f.o `test -f 'x17f.f' || echo './'`x17f.f > f90: Error: x17f.f, line 128: This name does not have a type, and must > have an explicit type. [RAND] > noise =3D rand(1) > --------------^ > make: *** [x17f.o] Error 1 I think these errors in the first and 17th examples are because those functions are not available in the f77 library for OSF1. lnblnk and especially rand are both problematic cross-platform. For example, you may have to use a different function with different name and arguments to get random numbers from fortran on OSF1. Alternatively, rand (and lnblnk) may be available to you only if you link in an extra "compatibility" library on OSF1. But if that fails, and you have to access some other fortran library functions on OSF1 to trim trailing blanks (lnblnk) and generate random numbers (rand), then hand-editing of examples x01f.fm4 and x17f.fm4 may be required to get those examples to work on OSF1. The Unresolved: pl_setcontlabelparam_ error for x09f is undoubtedly the second trailing underscore issue raising its head which I have discussed recently on list. Could you please look at your f77 options to see if you can turn on appending second underscores for names (such as pl_setcontlabelparam) with a single underscore already in them? The equivalent option for g77 is -f-second-underscore (the default which I use for that compiler since it is claimed to be most compatible with what is done for other Unix fortran compilers). > > And then all made demos runs OK under xwin Was this for plplot-5.2.1.cvs.20040104.tar.gz (which used AC_F77_LIBRARY_LDFLAGS) or from updated CVS from this morning after Rafael removed AC_F77_LIBRARY_LDFLAGS from configure.ac? In any case I don't think removing AC_F77_LIBRARY_LDFLAGS from configure.ac will do any harm for OSF1 f77 since you seemed to compile the fortran examples under 5.2.1 OK, and we definitely didn't use AC_F77_LIBRARY_LDFLAG= S then. But please answer the specific question above so we know whether in future we are free to permanently move to using AC_F77_LIBRARY_LDFLAGS for OSF1 f77. Alan __________________________ Alan W. Irwin email: ir...@be... phone: 250-727-2902 Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the PLplot scientific plotting software package (plplot.org), the Yorick front-end to PLplot (yplot.sf.net), the Loads of Linux Links project (loll.sf.net), and the Linux Brochure Project (lbproject.sf.net). __________________________ Linux-powered Science __________________________ |
From: <jc...@fe...> - 2004-01-06 00:18:41
|
On Monday 05 January 2004 20:47, Alan W. Irwin wrote: | On 2004-01-05 19:58-0000 Jo=E3o Cardoso wrote: | > On Monday 05 January 2004 06:19, Alan W. Irwin wrote: | > | On 2004-01-04 21:38-0800 Alan W. Irwin wrote: | > | > Because of the troubles we have encountered, we should also not | > | > just rely on the documentation that AC_F77_LIBRARY_LDFLAGS is | > | > supposed to work. In addition we should test that it works on as | > | > many Unix fortran platforms as possible. The SF compile farm is | > | > supposed to have a sparc8 processor available so I may try that in | > | > the next few days if they have the fortran compiler installed. | > | > Joao, will you try fortran on your OSF1 platform using the latest | > | > tarball? | > | > Yes. A quick test. =2E.. | Was this for plplot-5.2.1.cvs.20040104.tar.gz (which used yes. I will try to anwser the other questions latter. Joao |
From: Brian D. W. <bdw...@ph...> - 2003-12-28 05:25:45
Attachments:
plplot.spec
|
Hi Folks, I've updated the previous version of my Mandrake 9.2 spec file (attached) to work with Rafael's latest 12/27 CVS tarball. This time no patches are needed :) (as long as linuxvga is disabled). The file installation locations work fine using DESTDIR during the packaging "make install" step. The rpm list is libplplot9-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-devel-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-driver-gd-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-driver-gnome-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-drivers-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-java-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-static-devel-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-tcltk-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-tcltk-devel-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot9-tcltk-static-devel-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot-octave-5.2.1.cvs.20031227-1mdk.i586.rpm libplplot-python-5.2.1.cvs.20031227-1mdk.i586.rpm plplot-5.2.1.cvs.20031227-1mdk.i586.rpm plplot-tcltk-5.2.1.cvs.20031227-1mdk.i586.rpm I'd appreciate any more comments about the files distribution, package naming, etc. I've tried to use Mandrake's conventions as per their RPM-HOWTO. I suspect the same spec file would work for other RPM based distributions with minor modifications. For now the binary rpms can be downloaded by ftp from: ftp://keck.ucsf.edu/pub/urpmi/Mandrake/9.2/RPMS or using urpmi with the medium: urpmi.addmedia bdw_contrib_i586_9.2 ftp://keck.ucsf.edu/pub/urpmi/Mandrake/9.2/RPMS with ./hdlist.cz The source rpm is in ftp://keck.ucsf.edu/pub/urpmi/Mandrake/9.2/SRPMS After getting the blessing of the plplot-developers I'd like to submit this to Mandrake's Cooker. Cheers, Brian P.S. If you'd like to rebuild the rpms, I've packaged libcd (cgmdraw) and repackaged tcl/tk/itcl from Mandrake to add the needed development files. You can get these from the above ftp/urpmi site as cd-1.3-1mdk.i586.rpm libcd1-devel-1.3-1mdk.i586.rpm expect-8.4.2-3mdk.i586.rpm itcl-8.4.2-3mdk.i586.rpm tcl-8.4.2-3mdk.i586.rpm tcllib-8.4.2-3mdk.i586.rpm tclx-8.4.2-3mdk.i586.rpm tk-8.4.2-3mdk.i586.rpm On Sat, 27 Dec 2003, Rafael Laboissiere wrote: > I just released a new CVS snapshot distribution tarball for PLplot > (plplot-5.2.1.cvs.20031227.tar.gz). It has only four changes in relation to > the yesterday's tarball, but one of them can make a big difference for Mac > OS X users. Hopefully, both the "/lib/cpp" and the "AMDEP" problems are > fixed now. > > The tarball is available at the usual place: > > http://people.debian.org/~rafael/plplot.html > > Please, test and report. > > -- ====================================================================== Brian D. Wright Tel: (415)476-1007 Dept. of Physiology, Box 0444 Fax: (415)476-4929 Keck Center for Integrative Neuroscience bdw...@ph... University of California, San Francisco 513 Parnassus Avenue San Francisco, CA 94143-0444 ====================================================================== |