update giac spkg

[frederichan]

Update the spkg to upstream 1.2.2-95

The giac tarball built from spkg-src to put in upstream/ is there:

​http://webusers.imj-prg.fr/~frederic.han/xcas/sage/giac-1.2.2.95.tar.bz2

Switch to .bz2 as wished in #18826

add --disable-ao in spkg-install to avoid libao dependency. remove the --disable-gmpxx in spkg-install. The built seems not broken anymore.

updtate giacpy to 0.6 doctest fix + change of name to giacpy_sage. The tarball to put in upstream is: ​http://webusers.imj-prg.fr/~frederic.han/xcas/giacpy/sage/giacpy_sage-0.6.tar.gz

The modifications about giacpy are there: ​https://gitlab.math.univ-paris-diderot.fr/han/giacpy-sage

The 1.2.2.85 version of giac should fix a race condition in icas that could appear during the test suite.

[leif]

You're not (yet?) listed ​here, so hopefully I've not misspelled your name.

[frederichan]

Replying to leif:

You're not (yet?) listed ​here, so hopefully I've not misspelled your name.

its done.

I need to test more. Ex:

Failed example:
[giacpy-0.5.6]     f=1/(5+cos(4*x));f.int(x)
[giacpy-0.5.6] Expected:
[giacpy-0.5.6]     1/2/(2*sqrt(6))*(atan(2*tan(4*x/2)/sqrt(6))+pi*floor(4*x/2/pi+1/2))
[giacpy-0.5.6] Got:
[giacpy-0.5.6]     1/2/(2*sqrt(6))*(atan((-sqrt(6)*sin(4*x)+2*sin(4*x))/(sqrt(6)*cos(4*x)+sqrt(6)-2*cos(4*x)+2))+4*x/2)

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​600e7a6	update to giacpy 0.6. start to change the name from giacpy to giacpy_sage
​adae5f7	fixes changes of name for giacpy in src/sage/libs/giac.py
​88c2d0a	update SPKG.txt

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​c660391

change the name of the spkg giacpy to giacpy_sage

[frederichan]

Comment 2 is not a problem. This change was done on purpose in giac, both answer are correct. giacpy-0.6 doctests can deal with both answers.

I am waiting a bit more because I have found a bug in 1.2.2.67 ​http://xcas.e.ujf-grenoble.fr/XCAS/viewtopic.php?f=3&t=1731

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​eb2591b

update to giac-1.2.2-69

[fbissey]

Darn, I have to update my ebuild for Gentoo. Since we have a spkg-src already should i fix it to get lapack in sage like I did in sage-on-gentoo or should we wait a bit? I would also prefer if the tarball and the folder in which giacpy_sage unfold were matching the new package name.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ddfde45

change the tarball name of giacpy to giacpy_sage

[frederichan]

Replying to fbissey:

Darn, I have to update my ebuild for Gentoo. Since we have a spkg-src already should i fix it to get lapack in sage like I did in sage-on-gentoo or should we wait a bit?

Feel free to modify as you want, I have put it on a public branch for this. I have no news from parisse about a lapack change except the discussion on sage-devel.

I would also prefer if the tarball and the folder in which giacpy_sage unfold were matching the new package name.

should be done now

[fbissey]

Replying to frederichan:

Replying to fbissey:

Darn, I have to update my ebuild for Gentoo. Since we have a spkg-src already should i fix it to get lapack in sage like I did in sage-on-gentoo or should we wait a bit?

Feel free to modify as you want, I have put it on a public branch for this. I have no news from parisse about a lapack change except the discussion on sage-devel.

Yes I can do a simple thing that will work in sage as it works on gentoo since we have .pc files for blas/lapack. Probably won't work for parisse.

I would also prefer if the tarball and the folder in which giacpy_sage unfold were matching the new package name.

should be done now

Thank you.

[fbissey]

Hum I am unwilling to touch the spkg-src while I don't completely understand it. I would regenerate Makefile.in and co and probably wreak havoc in the current setup. Better ship it as it is now. If we really want lapack support we'll work it out later.

Actually the current configuration will work on OS X I think and it is ok not to disable it as the runtime involved comes with the OS.

What is the last bit of spkg-inst supposed to do?

# create the script_subdirectory for the pexpect giac interface by running a trivial instruction
sage -c "giac(1)"
if [ $? -ne 0 ]; then
    echo >&2 "Error running the pexpect interface of giac"
    exit 1
fi

Where is the directory supposed to be created? I cannot see anything new when I run ./sage -p giac with this branch. I was, possibly, expecting something in local/share/sage/ext but found nothing there. Running ./sage -c "giac(1)" manually afterwards didn't have an impact either.

[frederichan]

Replying to fbissey:

Hum I am unwilling to touch the spkg-src while I don't completely understand it. I would regenerate Makefile.in and co and probably wreak havoc in the current setup. Better ship it as it is now. If we really want lapack support we'll work it out later.

Actually the current configuration will work on OS X I think and it is ok not to disable it as the runtime involved comes with the OS.

What is the last bit of spkg-inst supposed to do?

# create the script_subdirectory for the pexpect giac interface by running a trivial instruction
sage -c "giac(1)"
if [ $? -ne 0 ]; then
    echo >&2 "Error running the pexpect interface of giac"
    exit 1
fi

Where is the directory supposed to be created? I cannot see anything new when I run ./sage -p giac with this branch. I was, possibly, expecting something in local/share/sage/ext but found nothing there. Running ./sage -c "giac(1)" manually afterwards didn't have an impact either.

Indeed It looks obsolete. may by since #17044. Moreover I think that it is the only dependency of SAGE_RUNTIMES cf #19672 comment 6 and jeroen answer. I am trying to remove both.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​912a9ce	Merge branch 'develop' of git://trac.sagemath.org/sage into giacupdate
​82bbbb4	Remove built of script_subdirectory for the pexpect interface. (Obsolete since #17044)
​95c7d7c	remove SAGE_RUNTIME dependencies because only sage -c "giac(1)" in spkg_install needed it. (Cf #19672)

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​7a407a5

add SAGERUNTIME in giacpy_sage dependencies

[jdemeyer]

The branch no longer applies.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​2805398	Merge branch 'develop' into public/giacupdate
​d2ae1dd	update giac to 1.2.2.81. adapt the cSolve patch

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​baa4eeb

update to giac-1.2.2.85. This version fix a race condition with icas and jordan that could appear with the test suite

[rws]

Passes make ptestlong with optional=atlas,ccache,giac,git_trac,mpir,python2,sage on a system where the old version was installed. I don't think there is something amiss in the package updates. Anyway it's an optional package and an improvement so I'll set positive.

[jdemeyer]

This conflicts with blocker ticket #21441, please rebase on top of that.

[git]

Branch pushed to git repo; I updated commit sha1. Last 10 new commits:

​bb613fd	update SPKG.txt
​0bf00dc	change the name of the spkg giacpy to giacpy_sage
​a6418e4	update to giac-1.2.2-69
​d601e2b	change the tarball name of giacpy to giacpy_sage
​6dcce70	Remove built of script_subdirectory for the pexpect interface. (Obsolete since #17044)
​f081cec	remove SAGE_RUNTIME dependencies because only sage -c "giac(1)" in spkg_install needed it. (Cf #19672)
​eb18297	add SAGERUNTIME in giacpy_sage dependencies
​1c01bc7	update giac to 1.2.2.81. adapt the cSolve patch
​d80f7a8	update to giac-1.2.2.85. This version fix a race condition with icas and jordan that could appear with the test suite
​a12a3ac	Merge branch 'public/giacupdate' of git://trac.sagemath.org/sage into giacupdate

[frederichan]

So I have rebased on u/jdemeyer/sage-pip-install

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​2c125a8	Merge branch 'master' of git://trac.sagemath.org/sage into giacupdatev2
​c709c1e	update to giac 1.2.2-95. bugfix version. Ex: (-1)^(2/7)

[fbissey]

I am not sure why this is not back to positive review. Let's try again.