Opened 10 years ago
Last modified 10 years ago
#11130 closed defect
Update PARI to version 2.5.0 — at Version 100
Reported by: | jdemeyer | Owned by: | jdemeyer |
---|---|---|---|
Priority: | critical | Milestone: | sage-4.8 |
Component: | packages: standard | Keywords: | pari spkg sd32 |
Cc: | mstreng, dimpase, jpflori | Merged in: | |
Authors: | Jeroen Demeyer, John Cremona | Reviewers: | John Cremona, Jeroen Demeyer |
Report Upstream: | N/A | Work issues: | |
Branch: | Commit: | ||
Dependencies: | #11230, #11234, #11321 (install this after building PARI) | Stopgaps: |
Description (last modified by )
We need to add bugfixes for
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1147 (#9334)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1153 (#10195)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1154 (#11604)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1184 (related to #9937)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1185 (PARI bug discovered by a Sage doctest)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1186 (PARI performance regression discovered by a Sage doctest)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1187 (PARI bug discovered by a Sage doctest)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1188 (PARI bug discovered by a Sage doctest)
- http://pari.math.u-bordeaux.fr/cgi-bin/bugreport.cgi?bug=1192 (#10767)
We will do this by updating to PARI version 2.5.0, which is equal to svn revision 13228.
See also:
- #9343, #10430 (previous PARI upgrades)
- #10247 (Sage 4.6 has PARI problems on Cygwin)
- #10240 (pari-2.4.3.svn-12577.p9 incorrectly checks for the shared library on Cygwin), included in the 2.4.3.alpha.p7 spkg at #11605
- #11605 (Typos in PARI's spkg-install (2.4.3.alpha.p5)), included in the 2.4.3.alpha.p7 spkg at #11605
Apply:
- spkg http://boxen.math.washington.edu/home/jdemeyer/spkg/pari-2.5.0.p2.spkg (jdemeyer)
- spkg http://boxen.math.washington.edu/home/jdemeyer/spkg/lcalc-1.23.p8.spkg (from #11321)
- 11130_sagelib.patch (jdemeyer, cremona, positive_review)
- 11130-sagelib-simon.patch (cremona)
- 11130_reviewer32.patch (jdemeyer)
- extcode patch 11130-extcode-simon.patch (cremona)
Change History (106)
comment:1 Changed 10 years ago by
- Description modified (diff)
- Summary changed from Add some patches to PARI to Update PARI to a more recent SVN version
comment:2 Changed 10 years ago by
- Description modified (diff)
- Status changed from new to needs_work
comment:3 Changed 10 years ago by
- Description modified (diff)
comment:4 Changed 10 years ago by
- Cc mstreng added
comment:5 Changed 10 years ago by
- Milestone changed from sage-4.7 to sage-4.7.1
comment:6 Changed 10 years ago by
- Description modified (diff)
comment:7 Changed 10 years ago by
- Description modified (diff)
comment:8 Changed 10 years ago by
- Description modified (diff)
comment:9 Changed 10 years ago by
- Description modified (diff)
comment:10 Changed 10 years ago by
- Description modified (diff)
comment:11 Changed 10 years ago by
- Description modified (diff)
comment:12 Changed 10 years ago by
- Dependencies set to #11230
- Description modified (diff)
comment:13 Changed 10 years ago by
- Description modified (diff)
comment:14 Changed 10 years ago by
- Description modified (diff)
comment:15 Changed 10 years ago by
- Cc dimpase added
comment:16 Changed 10 years ago by
- Description modified (diff)
comment:17 follow-up: ↓ 22 Changed 10 years ago by
comment:18 follow-up: ↓ 19 Changed 10 years ago by
After (1) Building a fresh 4.7.alpha5 from scratch, (2) installing the spkg lined here with "sage -f", (3) applying the patch here to the sage library and (4) running "sage -b", I find that the Sage build will not start up properly:
ImportError Traceback (most recent call last) ... /home/jec/sage-4.7.alpha5.11130/local/lib/python2.6/site-packages/sage/misc/functional.py in <module>() 36 37 ---> 38 from sage.rings.complex_double import CDF 39 from sage.rings.real_double import RDF, RealDoubleElement 40 /home/jec/sage-4.7.alpha5.11130/local/bin/gen.pxd in init sage.rings.complex_double (sage/rings/complex_double.c:15178)() ImportError: /home/jec/sage-4.7.alpha5.11130/local/lib/python2.6/site-packages/sage/libs/pari/gen.so: undefined symbol: defaultOut Error importing ipy_profile_sage - perhaps you should run %upgrade? WARNING: Loading of ipy_profile_sage failed.
so I cannot test anything.
Did I do something wrong?
comment:19 in reply to: ↑ 18 ; follow-up: ↓ 20 Changed 10 years ago by
comment:20 in reply to: ↑ 19 Changed 10 years ago by
comment:21 Changed 10 years ago by
- Dependencies changed from #11230 to #11230, #11234
- Description modified (diff)
- Owner changed from tbd to jdemeyer
comment:22 in reply to: ↑ 17 Changed 10 years ago by
Replying to cremona:
I am currently testing this. To avoid duplication of effort, I'll fix things I find in the elliptic_curves directory only.
Sounds good, I will look at everything else.
comment:23 Changed 10 years ago by
I added a patch which fixes the doctest failures in sage/rings/polynomial/polynomial_quotient_ring.py . In all cases the new output was equivalent to the old.
There are now just some failures in sage/schemes/elliptic_curves/ell_rational_field.py (2 doctests failed) and sage/schemes/elliptic_curves/ell_number_field.py (1 doctests failed).
comment:24 Changed 10 years ago by
- Reviewers set to John Cremona
- Status changed from needs_work to needs_review
ell_rational_field.py: only one failure, looks completely trivial but will be left until #11005 to be fixed.
ell_number_field.py: one failure, even more trivial: the final output is exactly the same as it was but the verbose output is a little different, so this can be easily cleaned up at #11005.
As far as I am concerned this needs *no more work* on this ticket, but I'll delay marking it positive review until the same is possible at #11005. Still, it certainly gets "needs review" on that basis.
comment:25 follow-up: ↓ 26 Changed 10 years ago by
This patch should also be tested on various systems first, PARI is known to produce different results on 32-bit and 64-bit systems.
comment:26 in reply to: ↑ 25 Changed 10 years ago by
Replying to jdemeyer:
This patch should also be tested on various systems first, PARI is known to produce different results on 32-bit and 64-bit systems.
Good point. I can test it on my 32-bit laptop.
comment:27 Changed 10 years ago by
- Description modified (diff)
- Work issues set to Test on various systems
comment:28 Changed 10 years ago by
- Description modified (diff)
comment:29 Changed 10 years ago by
- Description modified (diff)
comment:30 Changed 10 years ago by
- Work issues changed from Test on various systems to Test on various systems with #10247 applied
comment:31 Changed 10 years ago by
- Status changed from needs_review to needs_work
In sage/rings/number_field/number_field.py
, there is a test which fails only when -long
is not used.
comment:32 Changed 10 years ago by
New patch fixes the long time issue (by setting a random seed explicitly). I will test on a Mac OS X 10.4 32-bit PPC system.
comment:33 Changed 10 years ago by
comment:34 follow-up: ↓ 35 Changed 10 years ago by
Testing now on a 32-bit ubuntu system.
comment:35 in reply to: ↑ 34 Changed 10 years ago by
Replying to cremona:
Testing now on a 32-bit ubuntu system.
... which gave several more failures:
sage -t devel/sage-main/sage/rings/number_field/number_field.py # 1 doctests failed sage -t devel/sage-main/sage/schemes/elliptic_curves/ell_number_field.py # 2 doctests failed sage -t devel/sage-main/sage/schemes/elliptic_curves/ell_rational_field.py # 1 doctests failed sage -t devel/sage-main/sage/rings/number_field/number_field_element.pyx # 1 doctests failed sage -t devel/sage-main/sage/rings/polynomial/polynomial_quotient_ring.py # 3 doctests failed sage -t devel/sage-main/sage/rings/integer.pyx # 1 doctests failed
Details:
sage -t "devel/sage-main/sage/rings/number_field/number_field.py" ********************************************************************** File "/home/john/sage-4.7.alpha5/devel/sage-main/sage/rings/number_field/number_field.py", line 3616: sage: L.factor(a + 1) Expected: (Fractional ideal (1/2*a*b + a + 1/2)) * (Fractional ideal (-1/2*b - 1/2*a + 1)) Got: (Fractional ideal (-1/2*a*b - a - 1/2)) * (Fractional ideal (1/2*b + 1/2*a - 1))
is trivial,
sage -t "devel/sage-main/sage/rings/number_field/number_field_element.pyx" ********************************************************************** File "/home/john/sage-4.7.alpha5/devel/sage-main/sage/rings/number_field/number_field_element.pyx", line 1142: sage: t = (2*a + b)._rnfisnorm(L); t[1] Expected: (b - 2)*a + 2*b - 3 Got: (-4*b - 5)*a + 5*b + 6
needs looking at,
sage -t "devel/sage-main/sage/rings/polynomial/polynomial_quotient_ring.py"
are just like the ones I fixed before on the other machine, and
sage -t "devel/sage-main/sage/rings/integer.pyx" ********************************************************************** File "/home/john/sage-4.7.alpha5/devel/sage-main/sage/rings/integer.pyx", line 4389: sage: 7._bnfisnorm(CyclotomicField(7)) Expected: (-zeta7 + 1, 1) Got: (-zeta7^5 + zeta7^4, 1)
is also trivial (multiplying by a power of zeta does not change the norm!).
That's all I have time for today, so I hope things look similar on the other 32-bit system being tested.
comment:36 Changed 10 years ago by
- Description modified (diff)
I have added a new patch (which affects 4 files in sage/rings) after which all doctests pass on both 32- and 64-bit, apart from the simon-related ones.
That means that it is time to check #11005 on top of all this.
comment:37 Changed 10 years ago by
- Description modified (diff)
- Work issues Test on various systems with #10247 applied deleted
comment:38 Changed 10 years ago by
- Description modified (diff)
comment:39 Changed 10 years ago by
- Description modified (diff)
- Summary changed from Update PARI to a more recent SVN version to Update PARI to version 2.4.4.BETA
comment:40 Changed 10 years ago by
- Description modified (diff)
comment:41 follow-up: ↓ 42 Changed 10 years ago by
lcalc fails to build with the new PARI:
Lcommandline.cc: In function 'int main(int, char**)': Lcommandline.cc:476:54: error: 'allocatemoremem' was not declared in this scope Lcommandline.cc:403:16: warning: unused variable 'C' [-Wunused-variable] Lcommandline.cc:38:9: warning: unused variable 'i' [-Wunused-variable] In file included from ../include/L.h:537:0, from ../include/Lcommandline.h:36, from Lcommandline.cc:31: ../include/Lvalue.h: In member function 'Complex L_function<ttype>::find_delta(Complex, Double) [with ttype = int, Complex = std::complex<double>, Double = double]': Lcommandline.cc:463:50: instantiated from here ../include/Lvalue.h:37:21: warning: unused variable 'f2' [-Wunused-variable] make[1]: *** [Lcommandline.o] Error 1 make[1]: Leaving directory `/usr/local/src/sage-4.7.alpha5/spkg/build/lcalc-20100428-1.23.p6/src/src' make: *** [all] Error 2 Error building lcalc 'make'
comment:42 in reply to: ↑ 41 ; follow-up: ↓ 45 Changed 10 years ago by
Replying to jdemeyer:
lcalc fails to build with the new PARI:
Mike Rubenstein needs to be made aware of this.
comment:43 follow-up: ↓ 44 Changed 10 years ago by
What about genus2reduction? I am assuming John already made sure eclib is working.
comment:44 in reply to: ↑ 43 Changed 10 years ago by
Replying to fbissey:
What about genus2reduction? I am assuming John already made sure eclib is working.
A reasonable assumption, but...for a while now I have used Sage's Pari and NTL builds rather than building them myself independently, so I cannot yet confirm. (However, eclib's use of Pari is minimal, it's only for integer factorization.)
I am also preparing a new version of eclib, so I will combine these two tasks, though I don't have any time for this this week.
comment:45 in reply to: ↑ 42 Changed 10 years ago by
comment:46 Changed 10 years ago by
- Dependencies changed from #11230, #11234 to #11230, #11234, #11321
- Description modified (diff)
See #11321 for lcalc.
comment:47 Changed 10 years ago by
- Dependencies changed from #11230, #11234, #11321 to #11230, #11234, #11321 (install this '''after''' building PARI)
comment:48 follow-up: ↓ 49 Changed 10 years ago by
- Work issues set to Fix simon_two_descent doctests
I built Sage completely from scratch with the new PARI and lcalc packages, everything works. All tests pass except for the ones related to simon_two_descent.
comment:49 in reply to: ↑ 48 ; follow-up: ↓ 50 Changed 10 years ago by
Replying to jdemeyer:
I built Sage completely from scratch with the new PARI and lcalc packages, everything works. All tests pass except for the ones related to simon_two_descent.
So is the "Needs work" just because this ticket is dependent on #11005? I am planning to look now at the doctest failures relating to the Simon scripts and fix them, so that this ticket can then be closed, and the additional work needed on #11005 can depend on this.
comment:50 in reply to: ↑ 49 ; follow-up: ↓ 51 Changed 10 years ago by
- Description modified (diff)
comment:51 in reply to: ↑ 50 Changed 10 years ago by
Replying to jdemeyer:
Replying to cremona:
I am planning to look now at the doctest failures relating to the Simon scripts and fix them, so that this ticket can then be closed, and the additional work needed on #11005 can depend on this.
That's precisely what I proposed.
OK, so that's what I am working on doing. Patch up shortly. However, I ran into something awkward: the script ellQ.gp in data/extcode/pari/simon crashed randomly in the function ellredgen (using this new pari version). I saw that the offending lines had been changed in the newer ellQ.gp as found in #11005, so I made the same changes. This sorts the problem, but it means that for this ticket to be finished we need to also make a small patch to extcode. Any problems with that?
comment:52 Changed 10 years ago by
- Description modified (diff)
- Status changed from needs_work to needs_review
comment:53 Changed 10 years ago by
- Description modified (diff)
comment:54 Changed 10 years ago by
- Description modified (diff)
- Work issues Fix simon_two_descent doctests deleted
comment:55 Changed 10 years ago by
Is the only thing stopping this going forward the latest reviewer patch?
comment:56 follow-up: ↓ 57 Changed 10 years ago by
As far as I know, all patches and spkgs still need review. I can review your patches (but give me some time).
comment:57 in reply to: ↑ 56 Changed 10 years ago by
Replying to jdemeyer:
As far as I know, all patches and spkgs still need review. I can review your patches (but give me some time).
Well I considered that I had reviewed (positively) everything except my last patch fixing the Simon doctests so that's the only thing left needing a review (from you perhaps, especially as you made some extra fixes for 32/64 bit issues).
comment:58 Changed 10 years ago by
Also lcalc (#11321) needs review.
comment:59 Changed 10 years ago by
The part of the ticket that fixes #10240 is okay, I think.
comment:60 Changed 10 years ago by
- Description modified (diff)
- Status changed from needs_review to needs_work
- Summary changed from Update PARI to version 2.4.4.BETA to Update PARI to version 2.5.0
comment:61 Changed 10 years ago by
See #10240 for a request to merge that in 4.7.2 (or whatever the next version is) if this isn't getting merged quickly.
comment:62 Changed 10 years ago by
comment:63 Changed 10 years ago by
- Description modified (diff)
comment:64 Changed 10 years ago by
- Description modified (diff)
Needs to be rebased to sage-2.4.3.alpha7.spkg (#11605).
comment:65 Changed 10 years ago by
- Work issues set to Rebase on PARI(!) 2.4.3.alpha.p7 spkg from #11605
comment:66 Changed 10 years ago by
- Description modified (diff)
comment:67 Changed 10 years ago by
- Description modified (diff)
- Work issues Rebase on PARI(!) 2.4.3.alpha.p7 spkg from #11605 deleted
comment:68 Changed 10 years ago by
- Description modified (diff)
- Status changed from needs_work to needs_review
comment:69 Changed 10 years ago by
- Status changed from needs_review to needs_work
The spkg does not build on Mac OS X because of an upstream bug. There is a fix in PARI svn revision 13318 which I should add to this spkg.
comment:70 Changed 10 years ago by
My comment refered to a Mac OS X 10.4 PPC machine, other machines might still work.
comment:71 Changed 10 years ago by
- Description modified (diff)
- Status changed from needs_work to needs_review
New spkg fixing the OS X issue: http://boxen.math.washington.edu/home/jdemeyer/spkg/pari-2.5.0.p2.spkg
comment:72 Changed 10 years ago by
- Status changed from needs_review to needs_work
Never mind, still doesn't work. Need to also add revision 13330.
comment:73 Changed 10 years ago by
- Status changed from needs_work to needs_review
New version which does build on OS X 10.4, same location: http://boxen.math.washington.edu/home/jdemeyer/spkg/pari-2.5.0.p2.spkg
comment:74 Changed 10 years ago by
- Milestone changed from sage-4.7.1 to sage-4.7.2
- Priority changed from major to critical
I did a small rebase of 11130_sagelib.patch regarding sage/matrix/matrix2.pyx
comment:75 follow-up: ↓ 76 Changed 10 years ago by
- Status changed from needs_review to needs_info
Before I set about applying everything and testing, which I will do, can you tell me which Sage version I should be patching against? I have 4.7.1.rc1, is that the best? Or plain 4.7?
comment:76 in reply to: ↑ 75 ; follow-up: ↓ 77 Changed 10 years ago by
- Reviewers changed from John Cremona to John Cremona, Jeroen Demeyer
- Status changed from needs_info to needs_review
Replying to cremona:
Before I set about applying everything and testing, which I will do, can you tell me which Sage version I should be patching against? I have 4.7.1.rc1, is that the best? Or plain 4.7?
sage-4.7.1.rc1 for sure.
I am also preparing a Sage tarball with all spkgs and patches already applied, it should appear in http://boxen.math.washington.edu/home/jdemeyer/release/sage-4.7.1.rc1-pari-2.5.0/
comment:77 in reply to: ↑ 76 Changed 10 years ago by
Replying to jdemeyer:
Replying to cremona:
Before I set about applying everything and testing, which I will do, can you tell me which Sage version I should be patching against? I have 4.7.1.rc1, is that the best? Or plain 4.7?
sage-4.7.1.rc1 for sure.
OK, I am testing against that right now... John
I am also preparing a Sage tarball with all spkgs and patches already applied, it should appear in http://boxen.math.washington.edu/home/jdemeyer/release/sage-4.7.1.rc1-pari-2.5.0/
comment:78 follow-up: ↓ 81 Changed 10 years ago by
Starting with a clean 4.7.1.rc1 (I think) I successfully applied both spkgs (using SAGE_CHECK=yes) and 5 patches to sagelib and 1 to data/extcode, and did sage -ba.
Sage now starts up OK but I'm having a lot of problems with any more testing. For example, moving into devel/sage and doing "../../sage -tp 20 ./sage" results in very many errors, looks like in every file, like this:
Traceback (most recent call last): File "/home/jec/.sage//tmp/facade_sets.py", line 2, in <module> from sage.all_cmdline import *; File "/home/jec/sage-4.7.1.rc1/local/lib/python/site-packages/sage/all_cmdline.py", line 24, in <module> raise ValueError, msg ValueError: Attempted relative import in non-package
I checked (using bash's history) that I had remembered to hg qpush each patch.
Does this make any sense? I suppose I could rebuild 4.7.1.rc1 from scratch.
comment:79 follow-up: ↓ 80 Changed 10 years ago by
hen trying to install this on Cygwin, I get a lot of errors while trying to create the binary gp-2.5.exe along the lines of the following three types.
.../libpari.a(default.o): In function `sd_prompt': .../default.c:962: multiple definition of `_sd_prompt' gp.o:.../gp/gp.c:2359: first defined here
and
gp.o:gp.c:(.text+0x9e2): undefined reference to `_err_printf'
(lots of this kind of thing, for many functions), and
.../libpari.a(mp.o): In function `umodiu': .../kernel/gmp/mp.c::567: undefined reference to `__imp____gmpn_mod_1'
The latter type I've already seen while trying to ./sage -f the previous Pari (2.4.3.alpha.px), but I haven't seen the first kind or second kind before. Something similar to the third one I've seen in some build notes of William's as being a problem with gmp being -I'ed before some other include directories, and the first one looks like one of these "double inclusion" problems.
comment:80 in reply to: ↑ 79 ; follow-up: ↓ 82 Changed 10 years ago by
Replying to kcrisman:
When trying to install this on Cygwin, I get a lot of errors while trying to create the binary gp-2.5.exe along the lines of the following three types.
.../libpari.a(default.o): In function `sd_prompt': .../default.c:962: multiple definition of `_sd_prompt' gp.o:.../gp/gp.c:2359: first defined here
Looks as if you're somehow^{TM} mixing the old PARI (at least the static library libpari.a
) with the new one; PARI 2.5.0 doesn't have an sd_prompt()
function in default.c
at all.
and
gp.o:gp.c:(.text+0x9e2): undefined reference to `_err_printf'
... while this function doesn't exist in the older PARI.
(lots of this kind of thing, for many functions), and
.../libpari.a(mp.o): In function `umodiu': .../kernel/gmp/mp.c::567: undefined reference to `__imp____gmpn_mod_1'
The latter type I've already seen while trying to ./sage -f the previous Pari (2.4.3.alpha.px), but I haven't seen the first kind or second kind before. Something similar to the third one I've seen in some build notes of William's as being a problem with gmp being -I'ed before some other include directories, and the first one looks like one of these "double inclusion" problems.
comment:81 in reply to: ↑ 78 ; follow-up: ↓ 86 Changed 10 years ago by
Replying to cremona:
ValueError?: Attempted relative import in non-package
I have no idea why this happens. Can you try to build http://boxen.math.washington.edu/home/jdemeyer/release/sage-4.7.1.rc1-pari-2.5.0/sage-4.7.1.rc1-pari-2.5.0.tar from scratch and test that? If this works, probably you did something wrong in merging the spkgs/patches.
comment:82 in reply to: ↑ 80 ; follow-up: ↓ 83 Changed 10 years ago by
Looks as if you're somehow^{TM} mixing the old PARI (at least the static library
libpari.a
) with the new one; PARI 2.5.0 doesn't have ansd_prompt()
function indefault.c
at all.
Interestingly, it works fine from scratch (building)! Sorry for the noise, if noise it was.
Is there something about Pari that you shouldn't build it twice in the same location? I notice that it removes and replaces various things in $SAGE_LOCAL
before it's all done, so maybe that could happen. In principle, one should be able to ./sage -f
a new version, have something break, and then be able to ./sage -f
the original version and have everything fine. Or not even HAVE to ./sage -f
the original version. The fact that this doesn't work is not good - shouldn't we only copy things over/delete things after the build has successfully completed?
comment:83 in reply to: ↑ 82 ; follow-up: ↓ 85 Changed 10 years ago by
Replying to kcrisman:
Looks as if you're somehow^{TM} mixing the old PARI (at least the static library
libpari.a
) with the new one; PARI 2.5.0 doesn't have ansd_prompt()
function indefault.c
at all.Interestingly, it works fine from scratch (building)! Sorry for the noise, if noise it was.
Is there something about Pari that you shouldn't build it twice in the same location? I notice that it removes and replaces various things in
$SAGE_LOCAL
before it's all done, so maybe that could happen. In principle, one should be able to./sage -f
a new version, have something break, and then be able to./sage -f
the original version and have everything fine. Or not even HAVE to./sage -f
the original version. The fact that this doesn't work is not good - shouldn't we only copy things over/delete things after the build has successfully completed?
Well the gp executable is linked against libpari and there is a danger that the old one will be used when linking if you are not careful enough. It should be ok in principle but the precaution of removing previous libpari is a good one.
comment:84 Changed 10 years ago by
Just a quick question to confirm: do we need to apply the patch to extcode without worrying about #11005? It is not a dependency of this bug.
comment:85 in reply to: ↑ 83 Changed 10 years ago by
Replying to fbissey:
Replying to kcrisman:
Looks as if you're somehow^{TM} mixing the old PARI (at least the static library
libpari.a
) with the new one; PARI 2.5.0 doesn't have ansd_prompt()
function indefault.c
at all.Interestingly, it works fine from scratch (building)! Sorry for the noise, if noise it was.
Is there something about Pari that you shouldn't build it twice in the same location? I notice that it removes and replaces various things in
$SAGE_LOCAL
before it's all done, so maybe that could happen.
The PARI spkg doesn't delete anything of previous installs in advance. (The .alpha
.p5
though had a bug which caused the install script not to exit with an error if make
failed. Also, on Cygwin make -k
("keep going") was used, but that shouldn't matter here, and has been removed in the 2.5.0 spkgs.)
In principle, one should be able to
./sage -f
a new version, have something break, and then be able to./sage -f
the original version and have everything fine. Or not even HAVE to./sage -f
the original version.
Both should^{TM} be possible, at least unless the build succeeds but the actual installation fails at some point, in which case the former method should still work.
The fact that this doesn't work is not good - shouldn't we only copy things over/delete things after the build has successfully completed?
That's what we do.
Well the gp executable is linked against libpari and there is a danger that the old one will be used when linking if you are not careful enough. It should be ok in principle but the precaution of removing previous libpari is a good one.
Things are a little different on Cygwin, since the Sage Cygwin developers ;-) copy the DLL also to $SAGE_LOCAL/bin
such that gp
will find its shared library also there.
PARI (2.4.3) prepends its build path (with -L
) when using -lpari
, except for when it links gp
[-2.4
] "in place", but this happens after the new shared library (.so*
, *.dll
on Cygwin) has been copied and new symbolic links have been created. The new static library is built and copied later though, but that shouldn't matter unless gp
is statically linked on Cygwin, which shouldn't be the case.
If despite that something weird happens, this is IMHO limited to Cygwin; you may attach your spkg install logs ($SAGE_ROOT/spkg/logs/pari-*.log
) such that we can check this.
We also didn't have problems with upgrading to 2.4.3 btw., except that initially some dependencies of Sage's extension modules were missing such that those didn't get rebuilt in the first place (and used the old ones with a different version they had been linked to earlier).
comment:86 in reply to: ↑ 81 ; follow-up: ↓ 87 Changed 10 years ago by
Replying to jdemeyer:
Replying to cremona:
ValueError?: Attempted relative import in non-package
I have no idea why this happens. Can you try to build http://boxen.math.washington.edu/home/jdemeyer/release/sage-4.7.1.rc1-pari-2.5.0/sage-4.7.1.rc1-pari-2.5.0.tar from scratch and test that? If this works, probably you did something wrong in merging the spkgs/patches.
I'm downloading that now and will report back.
comment:87 in reply to: ↑ 86 ; follow-up: ↓ 88 Changed 10 years ago by
Replying to cremona:
Replying to jdemeyer:
Replying to cremona:
ValueError?: Attempted relative import in non-package
I have no idea why this happens. Can you try to build http://boxen.math.washington.edu/home/jdemeyer/release/sage-4.7.1.rc1-pari-2.5.0/sage-4.7.1.rc1-pari-2.5.0.tar from scratch and test that? If this works, probably you did something wrong in merging the spkgs/patches.
I'm downloading that now and will report back.
Built with no problems. But "make ptestlong" failed 100% just as before. The install and ptestlong logs are at http://www.warwick.ac.uk/staff/J.E.Cremona/ptestlong.log and http://www.warwick.ac.uk/staff/J.E.Cremona/install.log
comment:88 in reply to: ↑ 87 ; follow-up: ↓ 89 Changed 10 years ago by
Replying to cremona:
Built with no problems. But "make ptestlong" failed 100% just as before. The install and ptestlong logs are at http://www.warwick.ac.uk/staff/J.E.Cremona/ptestlong.log and http://www.warwick.ac.uk/staff/J.E.Cremona/install.log
The strange thing is that the traceback you get seems totally unrelated to PARI.
comment:89 in reply to: ↑ 88 Changed 10 years ago by
Replying to jdemeyer:
Replying to cremona:
Built with no problems. But "make ptestlong" failed 100% just as before. The install and ptestlong logs are at http://www.warwick.ac.uk/staff/J.E.Cremona/ptestlong.log and http://www.warwick.ac.uk/staff/J.E.Cremona/install.log
The strange thing is that the traceback you get seems totally unrelated to PARI.
I know, it is mad. After this happened with my 4.7.1.rc1 yesterday I wanted to review something else, so I did that using my 4.7.1.rc0 which was untouched, and found the exact same errors arising whenever I tried "sage -t". This is on a machine I have been using for Sage development and testing for a year, on which no-one but me installs stuff.
Here are some possibly relevant environment variables (not changed recently):
SAGE_PARALLEL_SPKG_BUILD=yes SHELL=/bin/bash LD_LIBRARY_PATH=:/usr/local/lib:/home/jec/sage-current/local/lib
The reason for the LD_LIBRARY_PATH is that I use Sage's built libraries for other work (libntl and libpari); but seeting that to just /usr/local/lib made no difference anyway.
comment:90 follow-up: ↓ 91 Changed 10 years ago by
Same happens with just ./sage -t ...
or make test[long]
, i.e. non-parallel?
We had strange errors with ptest[long]
or ./sage -tp N ...
when SAGE_PATH
was set (even to the empty string).
comment:91 in reply to: ↑ 90 ; follow-up: ↓ 92 Changed 10 years ago by
Replying to leif:
Same happens with just
./sage -t ...
ormake test[long]
, i.e. non-parallel?We had strange errors with
ptest[long]
or./sage -tp N ...
whenSAGE_PATH
was set (even to the empty string).
That is not set:
jec@fermat%env | grep SAGE SAGE_PARALLEL_SPKG_BUILD=yes
make testlong was no better, neither is testing a single file (e.g. ./sage -t devel/sage/sage/rings/infinity.py). Sage does start up properly.
Looking at the actual error message (why not?) I see that it complains that the line
from sage.all_cmdline import *;
which is presumably included in the testing framework, causes sage.all_cmdline.py to be imported, and then in line 24 of that file the line 24:
raise ValueError, msg
raises the error, with the message coming from a few lines above where there is a try/except block containing
from sage.all import * from sage.calculus.predefined import x preparser(on=True)
and this is causing some kind of circular import.
Anyway, by 4.7 build which I have been using for weeks, and which passed all the tests when I built it, now gives rise to exactly the same problem. So this is nothing to do with any of the code on this ticket, but instead something weird that has happened with my Sage installations.
Here's an idea: I have several background jobs running sage; is it possible that this is the problem? They will also be writing to ~/.sage. I'll try again when they have finished.
comment:92 in reply to: ↑ 91 Changed 10 years ago by
Replying to cremona:
Here's an idea: I have several background jobs running sage; is it possible that this is the problem? They will also be writing to ~/.sage. I'll try again when they have finished.
Well, easily checked:
$ env DOT_SAGE=/tmp ./sage -t devel/sage/sage/rings/infinity.py
comment:93 follow-up: ↓ 94 Changed 10 years ago by
considering the amount of broken test you have, does sage even start properly without wanting to test it.
comment:94 in reply to: ↑ 93 ; follow-up: ↓ 96 Changed 10 years ago by
Replying to fbissey:
considering the amount of broken test you have, does sage even start properly without wanting to test it.
Yes, all the version I have start and run fine, but none of them cab run any tests. I have never seen this before, but it's something to do with my setup (and several sage jobs running in the background on the same machine) so we should probably not fill up more of this ticket with it -- as long as someone else can do some testing!
comment:95 Changed 10 years ago by
Build against current 4.7.2_alpha0 on linux amd64. No problem that can be attributed to pari after running the long tests.
comment:96 in reply to: ↑ 94 Changed 10 years ago by
Replying to cremona:
Replying to fbissey:
considering the amount of broken test you have, does sage even start properly without wanting to test it.
Yes, all the version I have start and run fine, but none of them cab run any tests. I have never seen this before, but it's something to do with my setup (and several sage jobs running in the background on the same machine) so we should probably not fill up more of this ticket with it -- as long as someone else can do some testing!
By deleting most of my ~/.sage I successfully ran all the tests (i.e. make ptestlong completed successfully). This is on 64-bit ubuntu.
Sorry for the noise -- is there any reason not to give this a positive review and get it into the 4.7.2 series as soon as possible?
comment:97 Changed 10 years ago by
- Keywords sd32 added
comment:98 Changed 10 years ago by
- Cc jpflori added
comment:99 Changed 10 years ago by
One doctest failure with sage-4.7.2.alpha2, fix included in new 11130_sagelib.patch:
sage -t -long -force_lib devel/sage/sage/rings/number_field/number_field_ideal.py ********************************************************************** File "/mnt/usb1/scratch/jdemeyer/merger/sage-4.7.2.alpha2-pari-2.5.0/devel/sage-main/sage/rings/number_field/number_field_ideal.py", line 1097: sage: [P._S_ideal_class_log(S) for P in K.primes_above(11)] Expected: [[5, 2], [1, 1]] Got: [[3, 2], [3, 1]] **********************************************************************
comment:100 Changed 10 years ago by
- Description modified (diff)
Positive review for trac_11130-doctest-poly.patch and 11130_sagelib32.patch. To make things simpler, I have folded these two patches into my 11130_sagelib.patch (in any case, the commit message of one of your patches needed to be fixed). Together with John's earlier review of my 11130_sagelib.patch, this one patch now has positive review.
I am currently testing this. To avoid duplication of effort, I'll fix things I find in the elliptic_curves directory only.