New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Optional package p_group_cohomology-1.2 fails to install on Solaris 10 SPARC, as well as Debian Linux x86/x86_64 and MacOSX 10.5 (PPC) #8523
Comments
comment:1
this is not a Solaris-specific bug, but a general one, related to the failure to recognise that database_gap is installed. Please check out |
comment:2
Hi Dima! I saw at this ticket when it was created, but I understood from the ticket description that it was not possible to install database_gap. Are you saying that you can install database_gap, but the install script of the group cohomology package does not recognise it? What did you change to make it recognise database_gap? I am currently preparing a major upgrade of the group cohomology package. I expect it to be ready in about 2 weeks. Probably it would be easiest to incorporate the necessary changes in the new version. What do you think? Best regards, Simon |
comment:3
Replying to @simon-king-jena:
Hi Simon, no, it was a Solaris-only problem, on the other patforms it worked just fine.
well, when did it work last time, with which Sage release? So here is the diff:
well, the fix is trivial, and it's better to have a working version for the next release already... Dima |
comment:4
Replying to @dimpase:
Starting with an spkg, I installed it in sage-4.3.1 to the least. And calling spkg-install directly is what I do all the time, with the newest sage.
Thank you for the diff. I will change my spkg-install accordingly. Best regards, Simon |
comment:5
Replying to @simon-king-jena:
Simon, calling spkg-install directly is certainly NOT the way it is meant to be installed. Please check that this works for you now, too.
you better just grab the spkg linked above, and tell Minh (and/or the release manager) to upgrade the sagemath.org repository This way, I could give it a positive review (pretending you did it all:-)). Best,
|
comment:6
Replying to @dimpase:
I think we talk about totally different situations. You seem to talk about version 1.2, which is published and should certainly be installable by a user doing sage -i. But IMO, sage -i is the way to go only if the package is finished. I am talking about the yet-to-be-published version 2.0. I am still not finished with all details of the new algorithms and documentation, and it has not being packaged yet. I will certainly not do sage -i while developing new algorithms. Namely, before doing sage -i, one has to have a spkg. Thus, I would have to do sage -pkg after each tiny little change, 20-50 times a day! That's clumsy! Moreover, sage -i should be equivalent to unpacking the spkg (well, it is unpacked since I didn't pack my development version yet) and calling spkg-install (plus, perhaps, spkg-check) in the sage environment. So, if spkg-install works (which it does for me) then sage -i should work as well. So, testing whether sage -i still works will only be the last step before publishing version 2.0.
You seem to talk about version 1.2. This version did work, on quite a broad range of platforms. I never had the problem that you describe. In fact, I just tested (sage 4.3.4 on Opensuse; you said that the problem is not platform specific), and sage -f p_group_cohomology-1.2.spkg (without your changes) came easily beyond the point where the existence of database_gap is tested. Then, I interrupted with Ctrl-C. So, can you please tell me how I can reproduce the problem that you met?
I don't plan to re-publish version 1.2, unless I can reproduce the problem. But I will pull it into version 2.0.
WHAT? Sorry for shouting, but certainly the repository was not off. I don't know if you ever did sage -pkg, but it gives an unmistakable warning if the repository is not fine. That you had to do hg commit is clear. But certainly spkg-install was in the repository. So, why hg add?
Here I am not so sure, but I thought that I did not include src/db in the repository, for this reason. BTW, the online database does not contain the cohomology rings for the groups of order 64 -- these are only provided by the database in the package.
Now I am totally confused. First of all, I think that this ticket is a "wontfix", because it will soon be superseded by another ticket that I will open when I publish version 2.0. Moreover, it is about a problem that I can not reproduce. Then you say you will give a positive review -- to changes that you did yourself? Best regards, Simon |
comment:7
Concerning the mercurial repository: I just verified that SPKG.txt and spkg-install are tracked in the spkg's repository, and that the database for the groups of order 64 is not tracked. So, I can not reproduce any of the issues that you are reporting. |
comment:8
Replying to @simon-king-jena:
Simon, In http://sagemath.org/packages/optional/p_group_cohomology-1.2.spkg
the "?" status means something like "hg does not know about this file; it is there, but I have no clue how to deal with it" which is obviously not what one likes to have, even if sage -spkg
Probably it slipped through when the package was reviewed. Best, |
comment:9
Replying to @simon-king-jena:
Sure, I do talk about 1.2 (and I was baffled by what you wrote). And 1.2 is broken, it does not install on any Sage 4.3.4 I have (as I was busy with updating gap-related spkgs and cvxopt spkg, I have a large supply presently :-)).
ok, so SmallGroups are there, but I cannot install the package (I just copied the canonical 1.2 version to /tmp)
Please see above. I don't have access to SUSE systems, but it fails on a range of Ubuntu and Debian Linuxes, as well as on Solaris (Sparc) and on MacOSX 10.5 (PPC), all of these with Sage 4.3.4.
Try it on sage.math or boxen.math and see for yourself, if you like.
as I explained in another reply, to bring it in line to what you wrote in SPKG.txt.
OK, sorry about this.
My apologies. As usual, the joke is lost in electronic communications... I meant to say I explained it to you all, leaving you almost nothing to figure out, so it would be a breeze to give it a positive review.
well, 2.0 is still at least weeks away, and then reviewing, etc etc. Meanwhile there is a broken spkg on the list of optional packages, and the fix is ready. So, please, please, let us fix it, and be done with. Best, |
comment:10
Replying to @dimpase:
In http://sagemath.org/packages/optional/p_group_cohomology-1.2.spkg (do you talk about the same repository?) hg reports
Ah, now I understand. In fact, these files do not belong into the hg repository, IMO. The On the other hand, look into the Changes to third party code should be made transparent, and I thought that if there are many changes then the resulting large diff file would be less transparent than having both versions simultaneously: The version from third party, and our modified version. Only the code in
Exactly. And it also says:
I thought this is clear enough. IIRC, the developer's guide says that the Here, we have original code, and therefore Probably I should better add a Best regards, Simon |
comment:11
Replying to @dimpase: Sure, I do talk about 1.2 (and I was baffled by what you wrote). And 1.2 is broken, it does not install on any Sage 4.3.4 I have It did compile on several platforms. Personally, I tested Suse linux (two different machines at my university), sage-math, and Intel mac (Darwin). David Joyner and William Stein tested it on various other platforms, like Ubuntu. The issues found there had been sorted out. For instance, one compiled from source on boxen.math OK, then I will try to build sage on boxen and see what happens. I'll reply later what happens. Please see above. I don't have access to SUSE systems, but it fails on a range of Ubuntu and Debian Linuxes, as well as on Solaris (Sparc) and on MacOSX 10.5 (PPC), all of these with Sage 4.3.4. OK, these had been tested with previous sage versions. So, now the picture becomes clearer: The package became broken by upgrading sage, and I think this is what you said.
See my reply there: These files, I think, do not belong under version control, but making this clear by .hgignore might be a good idea.
Right, that makes sense. So, I will test on boxen. If successful, I will probably then do the following:
The question is, though: Should this still be p_group_cohomology-1.2, or better p_group_cohomology-1.2.1? And who is giving a positive review? :) Best regards, Simon |
comment:12
PS: Sorry for the strange layout of my previous post. I just see that your post is not marked as a quote. It looked fine in "wysywig". |
comment:13
Replying to @simon-king-jena:
Right. Indeed, the most sensible thing seems to have .hgignore as follows:
(and do not forget to "hg add .hgignore", so that "hg status" would not report anything at all) OK, so my complaint was about "hg status" reporting lots of "?"-marked files, I didn't Best, Dima
|
comment:14
Replying to @simon-king-jena:
Apparently since the release of your spkg there was a semantics change in some Sage scripts that now causes us this headache.
there is a ready binary install for sage.math, you can just grab it instead, and try out on sage.math (not on boxen)
Right.
p_group_cohomology-1.2.p0 seems to be the proper name.
That's a tough one :-) Best, Dima
|
comment:15
Hi Dima! Now I am even more puzzled. I did precisely the following steps:
It installed fine, there was no error. I think this is how a user is supposed to install things, and so I can still not reproduce the error. Cheers, Simon |
comment:16
PS: Before using the package, I had to do gap_reset_workspace(), quit sage, start it again, and then I did successfully compute some cohomology rings. |
comment:17
Replying to @simon-king-jena: Hi,
I can confirm that indeed this way it works for me (on boxen with a compiled Sage 4.3.4-version), too. OK, so it appears to be an inconsistency in the Sage behaviour, when the option -i (or -f) is used to install the package. (AFAIK, these options are not something for experts only, they should work just the same...) It is, perhaps, a bug in newest_version script. Dima
|
comment:18
Replying to @dimpase:
That's a big relief!
But if one of the two usual ways of installing the package does work, then it might be acceptable to leave it as it is, and then to take care that both ways of installation work for p_group_cohomology-2.0. What do you think? |
comment:19
Replying to @simon-king-jena:
I spent an hour trying to sort this out, and my only conclusion is that on a clean and fresh install things work both ways. I propose to mark this ticket as "won't fix", as this is certainly nothing wrong Dima
|
comment:20
I think the problem is this: if you install "database_gap-4.4.12" by downloading the spkg file and typing
then it is installed, but the spkg file is not copied to SAGE_ROOT/spkg/optional or to SAGE_ROOT/spkg/standard. There is a placeholder file in SAGE_ROOT/spkg/installed, and this is probably what you should look for to test whether it's installed -- this is what the file SAGE_ROOT/local/bin/sage-spkg does. Note that after this, if you run "install_package('database_gap')" or "sage -i database_gap-4.4.12.spkg", it says that it's installed. (If you run "install_package('database_gap')", on the other hand, then Sage downloads the spkg file and puts it in spkg/optional, so your spkg-install file finds it.) |
comment:21
Now that's weird. I took the spkg-install as you suggested. But the test for the presence of database_gap failed:
In other word: For me, the original version works, but the modified version doesn't. Best regards, Simon |
comment:22
Replying to @jhpalmieri:
... and this is what I did in my spkg-install (for exactly the reason you explain):
Cheers, Simon |
comment:23
I have a file "database_gap-4.4.12" in SAGE_ROOT/spkg/installed, and I just did this:
and it didn't find it. It looks to me like "newest_version" searches for a file ending in ".spkg", and the files in spkg/installed don't match this. The script SAGE_ROOT/local/bin/sage-spkg does something else to check for installation:
where $INSTALLED is SAGE_ROOT/spkg/installed, $PKG_NAME is the name without "spkg", and you could omit "-a $FORCE", I think. |
comment:24
Replying to @jhpalmieri:
Good, I'll check this. What are the opinions: Can this wait until p_group_cohomology-2.0 is finished (since version 1.2 does work, provided the users install everything the "normal" way)? Or should there been a version 1.2.p0? Best regards, Simon |
comment:25
Hi John! I don't know how it works what you suggested. First of all, $INSTALLED has no value in the sage shell:
Could it be that it is only defined inside And one problem with this method is that (if I understand correctly) one has to provide the exact version of the package whose existence is being tested. But I don't want a method to test for the presence of database_gap-4.4.12 (so that it wouldn't work if database_gap-4.4.10 is installed). I want to test if any version of database_gap is there. And I think that this is exactly the purpose of the script Cheers, Simon |
comment:26
Replying to @simon-king-jena: OK, so it seems that I provided a "fix" that does not always work from within Sage, and your original package setup In your case, the safest would be just to call "sage -gap" and see if it has SmallGroups stuff loaded.
is nonempty. Best, |
comment:27
Replying to @dimpase:
This is great! It is an ideal solution, in the sense that it tests exactly what is needed. Namely, we actually don't care about the database_gap spkg -- it is alright if the user got his SmallGroups library from a different source. The change is little. So, I guess I'll create a version 1.2.p0, adding .hgignore, mentioning you in SPKG.txt (I am rewriting it in reverse chronological order, by the way), and change spkg-install; actually, I already tested that it works on my computer. Probably, if you or Dave are able to install the package with the new spkg-install, then you are entitled to give a positive review, so that the release manager can put it into the repository of optional packages. Thank you very much! Simon |
comment:28
Replying to @simon-king-jena:
That's right.
I think that's right, too. I don't know much about writing shell scripts, but this seems to work:
($SAGE_PACKAGES is defined by running "sage -sh".) This script defines PKG_BASE to be the name of the package without the version number: everything up to the first hyphen or period, whichever comes first. |
comment:29
Hi! I created a new patch level version, which I was able to install on sage.math by
Changes:
The test for the presence of Small Groups is now
Note that this is really a test for the Small Groups library, rather than for the database_gap spkg. The error message is a bit more descriptive as well. hg status does not return question marks. So, ready for review! Cheers, Simon |
comment:30
Replying to @simon-king-jena: OK, good, I checked that it also works on boxen and Solaris (Sparc)(t2).
I give it a positive review (despite my name mentioned in SPKG, I only contributed an idea for a fix, but I didn't touch the code). Best, |
Author: Simon King |
Changed keywords from none to recognise database_gap, group cohomology |
comment:31
Hi Dima! Fine! Then I change the "keywords", "author" and "reviewer" fields accordingly. Is it needed to notify a release manager, or are these guys reading any ticket anyway? Best regards, |
Reviewer: Dima Pasechnik |
comment:32
Replying to @simon-king-jena:
IMHO they check at least the positively reviewed ones... Dima |
comment:33
Replying to @dimpase:
OK. To the release manager (for avoiding confusion): The positive review is for the package posted at http://sage.math.washington.edu/home/SimonKing/Cohomology/p_group_cohomology-1.2.p0.spkg Simon |
comment:34
Merged 2010/04/20. |
Hardware & associated software
== Sage version ==
== The problem with the optional p_group_cohomology-1.2 ==
This looks like being related to #8514, since p_group_cohomology-1.2 needs the gap database. But #8514 seems to be a mess, and might not be easy to sort out.
CC: @simon-king-jena
Component: packages: optional
Keywords: recognise database_gap, group cohomology
Author: Simon King
Reviewer: Dima Pasechnik
Issue created by migration from https://trac.sagemath.org/ticket/8523
The text was updated successfully, but these errors were encountered: