Groups keyboard shortcuts have been updated
Dismiss
See shortcuts
Last steps before releasing v1.4.0
34 views
Skip to first unread message
Clement Pernet
Oct 9, 2015, 4:05:50 AM10/9/15
to linbox...@googlegroups.com
LinBoxers,
I'd like to cut the release 1.4.0 in the very near future (say in the coming weeks).
The goal is not to have the whole set of changes that we envisionned since years for a LinBox-2.0
implemented, but to be able to distribute the library in its more or less current state, that still
has a vast amount of improvements compared to the previous release 1.3.2 (from 2012).
We've promised such a release since already too long to users, in particular the SageMath community,
and it has become a blocker for several related open tickets there.
Not to mention that the current version of LinBox in Sage is frustratingly slow and makes a pretty
bad publicity of our software.
This email is to ask each one of you, developper,
- what is the status of your "work in progress" code in LinBox
- can you list the bits of it that you find of big enough relevance, to be release in 1.4.0, and
that you can reasonably ensure to finish and push in within a week or two.
- for code that you recently pushed in LinBox, can you make sure that you have a proper test-suite,
example files, and possibly a benchmark-file.
In case the test-suite reveals bugs that are too long to fix, then disable this piece of code, and
postpone the fix to the next release.
Once again, the goal is not to have a feature complete release but one, that works.
Information on the current status:
- most development action has happened in the Givaro and fflas-ffpack library recently. Both were
recently released, and I plan to cut a new givaro-4.0.1 and fflas-ffpack-2.2.0 at the same time as
linbox-1.4.0 to integrate the recent fixes to the ecosystem
- linbox test-suite (make check) passes, and the linbox-sage library compiles and works in sage (the
full sage testsuite passes with the new versions of the givaro-fflas-ffpack-linbox ecosystem).
- linbox full test-suite (make full-check) has some fails. In particular test-givaro-zpz
test-givaro-zpzuns fail due to pb with the enveloppe/archetype systems (See issue #7).
Best.
Clément
I'd like to cut the release 1.4.0 in the very near future (say in the coming weeks).
The goal is not to have the whole set of changes that we envisionned since years for a LinBox-2.0
implemented, but to be able to distribute the library in its more or less current state, that still
has a vast amount of improvements compared to the previous release 1.3.2 (from 2012).
We've promised such a release since already too long to users, in particular the SageMath community,
and it has become a blocker for several related open tickets there.
Not to mention that the current version of LinBox in Sage is frustratingly slow and makes a pretty
bad publicity of our software.
This email is to ask each one of you, developper,
- what is the status of your "work in progress" code in LinBox
- can you list the bits of it that you find of big enough relevance, to be release in 1.4.0, and
that you can reasonably ensure to finish and push in within a week or two.
- for code that you recently pushed in LinBox, can you make sure that you have a proper test-suite,
example files, and possibly a benchmark-file.
In case the test-suite reveals bugs that are too long to fix, then disable this piece of code, and
postpone the fix to the next release.
Once again, the goal is not to have a feature complete release but one, that works.
Information on the current status:
- most development action has happened in the Givaro and fflas-ffpack library recently. Both were
recently released, and I plan to cut a new givaro-4.0.1 and fflas-ffpack-2.2.0 at the same time as
linbox-1.4.0 to integrate the recent fixes to the ecosystem
- linbox test-suite (make check) passes, and the linbox-sage library compiles and works in sage (the
full sage testsuite passes with the new versions of the givaro-fflas-ffpack-linbox ecosystem).
- linbox full test-suite (make full-check) has some fails. In particular test-givaro-zpz
test-givaro-zpzuns fail due to pb with the enveloppe/archetype systems (See issue #7).
Best.
Clément
B Saunders
Oct 9, 2015, 5:09:54 PM10/9/15
to linbox...@googlegroups.com
LinBoxers,
Below is rearranged and annotated output of fullcheck. A lot doesn't work because of recent extensive structural changes and because of areas of neglect. I made 5 categories, namely:
1. I say move to attic.
2. Move to attic?
3. Keep, but exclude from fullcheck
4. Needs work
5. OK to release as skipped in fullcheck
Barring objection, I will move the first two categories to the attic.Below is rearranged and annotated output of fullcheck. A lot doesn't work because of recent extensive structural changes and because of areas of neglect. I made 5 categories, namely:
1. I say move to attic.
2. Move to attic?
3. Keep, but exclude from fullcheck
4. Needs work
5. OK to release as skipped in fullcheck
I say move to attic. OK?
test-block-ring skipped, non commutative rings not supported
test-dense-zero-one skipped, half baked, bds responsible
test-lidia-gfq skipped, Lidia not present
test-ffpack skipped, testTURBO fails, move to ffpack tests?
test-ftrmm skipped, should move to fflas tests?
test-image-field skipped, deprecated
test-modular skipped, deprecated
test-modular-byte skipped, deprecated
test-modular-short skipped, deprecated
test-quad-matrix skipped, depends on out-of-date blackbox/zo.h
Move to attic?
test-givaro-zpz build OK, run OK. tested in Givaro?
test-givaro-zpzuns build OK, run OK. tested in Givaro?
test-rat-minpoly skipped, stale test. solns over QQ need fresh tests
test-rat-solve skipped, stale test. solns over QQ need fresh tests
test-sparse skipped, superceded by test-sparse2
Keep, but exclude from fullcheck
test-la-block-lanczos skipped, not maintained. operator >> missing
test-mg-block-lanczos skipped, not maintained
test-sparse-map-map skipped, curious use of nonexistant next() in Extension
Needs work
test-blas-domain build FAIL
*test-blas-matrix build OK, run OK. intermittently fails
test-charpoly build FAIL
test-dyadic-to-rational build FAIL
test-fibb build OK, run FAIL
test-gf2 build OK, run OK. not much is tested there
test-last-invariant-factor build FAIL
test-matrix-domain build OK, run OK. intermittent row permutation failure
test-modular-balanced-double skipped, test and modular-balanced disagree on init/convert
test-modular-balanced-int skipped, test and modular-balanced disagree on init/convert
test-moore-penrose skipped, inf loop
test-ntl-rr build OK, run FAIL
test-ntl-zz_p build FAIL
test-order-basis build OK, run FAIL
test-poly-det build FAIL
test-qlup build OK, run FAIL
*test-rank-u32 skipped, intermittent error
*test-rat-charpoly skipped, inf loop
test-rational-solver-adaptive build FAIL
test-rational-solver build FAIL
test-smith-form-adaptive build FAIL
test-smith-form-binary build FAIL
test-smith-form build FAIL
test-smith-form-iliopoulos build FAIL
test-solve build OK, run OK. most of the tests are commented out
test-solve-nonsingular skipped, BY responsible
test-toom-cook build FAIL
OK to release as skipped in fullcheck
test-opencl-domain skipped, OpenCL not present
test-optimization skipped, not in test form
test-tutorial skipped, incomplete test
--
You received this message because you are subscribed to the Google Groups "linbox-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linbox-devel...@googlegroups.com.
To post to this group, send email to linbox...@googlegroups.com.
Visit this group at http://groups.google.com/group/linbox-devel.
For more options, visit https://groups.google.com/d/optout.
B Saunders
Oct 19, 2015, 10:45:04 AM10/19/15
to linbox...@googlegroups.com
Responding to Clement's note on a 1.4 release, my goals include:
1. Get tests working, including some issues I posted at github about rank computation, and including also some rational reconstruction that has mysteriously broken, breaking the rat solves.On Fri, Oct 9, 2015 at 4:05 AM, Clement Pernet <Clement...@gmail.com> wrote:
Clement Pernet
Oct 19, 2015, 11:07:14 AM10/19/15
to linbox...@googlegroups.com
Hi,
Thanks Dave for orgnazing this.
I agree on almost every point you propose.
Let me comment on those where I either disagree or that I have fixed:
> I say move to attic. OK?
test kind of make sense though (even if it is redundant with fflas-ffpack test-suite). I wouldn't
mind keeping it here, as a fullcheck target (even maybe a basic check target), as long as I did not
sit down and made sure that all tests in it are already done in fflas-ffpack test-suite.
> test-ftrmm skipped, should move to fflas tests?
It was already moved and improved. It should definitely be moved to the attic
> Move to attic?
> test-givaro-zpz build OK, run OK. tested in Givaro?
> test-givaro-zpzuns build OK, run OK. tested in Givaro?
These 2 tests should stay for they test the integration of these Givaro fields in LinBox. In
particular whether the archetype/enveloppe mechanism works on them.
This is currently broken (issue #7: https://github.com/linbox-team/linbox/issues/7 )
So it should be moved to category "Needs work"
> test-rat-minpoly skipped, stale test. solns over QQ need fresh tests
> test-rat-solve skipped, stale test. solns over QQ need fresh tests
The whole QQ part of LinBox is in a terrible state: either badly tested or tests fail or loop
infinitly. Let's try to fix it or disable it for the 1.4.0 release.
>8 years, so it's probably going to stay there unfinished and confuse users.
>
> Needs work
> test-blas-domain build FAIL
I've fixed it.
> *test-blas-matrix build OK, run OK. intermittently fails
I've fixed it.
> test-charpoly build FAIL
It does not fail on my side. Do you confirm it is fixed for you?
> test-dyadic-to-rational build FAIL
> test-fibb build OK, run FAIL
> test-gf2 build OK, run OK. not much is tested there
> test-last-invariant-factor build FAIL
Not seing any FAIL on my side.
> test-matrix-domain build OK, run OK. intermittent row permutation failure
Should be fixed now. Doesn't fail on my side.
> test-modular-balanced-double skipped, test and modular-balanced disagree on init/convert
> test-modular-balanced-int skipped, test and modular-balanced disagree on init/convert
> test-moore-penrose skipped, inf loop
> test-ntl-rr build OK, run FAIL
> test-ntl-zz_p build FAIL
This one always passes on my side.
> test-order-basis build OK, run FAIL
This one passes on my side (but with unwanted verbosity, Pascal, can you fix it?).
> test-poly-det build FAIL
This test looks totally unfinished (call to a missing function, missing license info, etc).
If this is work in progress that was never finished, we should move to the attic.
> test-qlup build OK, run FAIL
This is issue #10. JG, I added you as assignee, as it is mostly about code that you wrote.
> *test-rank-u32 skipped, intermittent error
Never seen it fail.
> *test-rat-charpoly skipped, inf loop
See my remark on QQ above. This one should be treated as test-rat-minpoly.
Clément
Thanks Dave for orgnazing this.
I agree on almost every point you propose.
Let me comment on those where I either disagree or that I have fixed:
> I say move to attic. OK?
> test-ffpack skipped, testTURBO fails, move to ffpack tests?
I fixed the TURBO test pb (TURBO is deprecated and moved to fflas-ffpack attic). But the rest of the
test kind of make sense though (even if it is redundant with fflas-ffpack test-suite). I wouldn't
mind keeping it here, as a fullcheck target (even maybe a basic check target), as long as I did not
sit down and made sure that all tests in it are already done in fflas-ffpack test-suite.
> test-ftrmm skipped, should move to fflas tests?
> Move to attic?
> test-givaro-zpz build OK, run OK. tested in Givaro?
> test-givaro-zpzuns build OK, run OK. tested in Givaro?
particular whether the archetype/enveloppe mechanism works on them.
This is currently broken (issue #7: https://github.com/linbox-team/linbox/issues/7 )
So it should be moved to category "Needs work"
> test-rat-minpoly skipped, stale test. solns over QQ need fresh tests
> test-rat-solve skipped, stale test. solns over QQ need fresh tests
infinitly. Let's try to fix it or disable it for the 1.4.0 release.
>
> Keep, but exclude from fullcheck
> test-la-block-lanczos skipped, not maintained. operator >> missing
> test-mg-block-lanczos skipped, not maintained
> test-sparse-map-map skipped, curious use of nonexistant next() in Extension
Ok. I wouldn't mind moving the whole block-lanczos to the attic. Nobody stepped in to fix it since
> Keep, but exclude from fullcheck
> test-la-block-lanczos skipped, not maintained. operator >> missing
> test-mg-block-lanczos skipped, not maintained
> test-sparse-map-map skipped, curious use of nonexistant next() in Extension
>8 years, so it's probably going to stay there unfinished and confuse users.
>
> Needs work
> test-blas-domain build FAIL
> *test-blas-matrix build OK, run OK. intermittently fails
> test-charpoly build FAIL
It does not fail on my side. Do you confirm it is fixed for you?
> test-dyadic-to-rational build FAIL
> test-fibb build OK, run FAIL
> test-gf2 build OK, run OK. not much is tested there
> test-last-invariant-factor build FAIL
> test-matrix-domain build OK, run OK. intermittent row permutation failure
> test-modular-balanced-double skipped, test and modular-balanced disagree on init/convert
> test-modular-balanced-int skipped, test and modular-balanced disagree on init/convert
> test-moore-penrose skipped, inf loop
> test-ntl-rr build OK, run FAIL
> test-ntl-zz_p build FAIL
> test-order-basis build OK, run FAIL
> test-poly-det build FAIL
This test looks totally unfinished (call to a missing function, missing license info, etc).
If this is work in progress that was never finished, we should move to the attic.
> test-qlup build OK, run FAIL
> *test-rank-u32 skipped, intermittent error
Never seen it fail.
> *test-rat-charpoly skipped, inf loop
See my remark on QQ above. This one should be treated as test-rat-minpoly.
Clément
Clement Pernet
Oct 19, 2015, 11:12:57 AM10/19/15
to linbox...@googlegroups.com
Ok, Excellent.
Good to know that you'll be working on the fibbs.
Recall that we shouldn't aim for new features for this 1.4.0 release. Mostly fix-up and clean-up.
Hence I'm not sure having a support for m4ri fibbs is a priority right now.
I'll definitely help solving the current open issues, and look at the QQ bugs.
Clément
> linbox-devel...@googlegroups.com <mailto:linbox-devel%2Bunsu...@googlegroups.com>.
Good to know that you'll be working on the fibbs.
Recall that we shouldn't aim for new features for this 1.4.0 release. Mostly fix-up and clean-up.
Hence I'm not sure having a support for m4ri fibbs is a priority right now.
I'll definitely help solving the current open issues, and look at the QQ bugs.
Clément
> To post to this group, send email to linbox...@googlegroups.com
> <mailto:linbox...@googlegroups.com>.
> Visit this group at http://groups.google.com/group/linbox-devel.
> For more options, visit https://groups.google.com/d/optout.
>
>
> For more options, visit https://groups.google.com/d/optout.
>
>
> --
> You received this message because you are subscribed to the Google Groups "linbox-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
> linbox-devel...@googlegroups.com <mailto:linbox-devel...@googlegroups.com>.
> You received this message because you are subscribed to the Google Groups "linbox-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
> To post to this group, send email to linbox...@googlegroups.com
> <mailto:linbox...@googlegroups.com>.
Reply all
Reply to author
Forward
0 new messages