-
-
Notifications
You must be signed in to change notification settings - Fork 491
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
Update ECL to the latest 9.8.4, needed for Solaris. #6564
Comments
comment:1
Creating an empty file $SAGE_HOME/local/include/ecl/config.h does allow the build of ECL to progress further, but it does not result in a working ecl installation. The build of ecl still fails. Whether that is a result of this problem I do not know, but I suspect not. Hence that error will be reported as another trac ticket. |
Attachment: install.log install.log showing ecl failure due to missing config.h |
comment:2
The attached install.log was created on 't2', a Sun T5240, with Sage-4.1.1.alpha0, whereas the first output I showed was created on my own Sun Blade 2000 with Sage-4.1. (Both used the same version of Solaris) (Note the Blade 2000 took about half the time of the T5240 to do this task), which shows the T5240 is not ideally suited to the sort of things currently being used for, but with more users, its performance will shine.) I see a few warnings - see below. What is clear is that the build often complains about things redefined in {{{
Anyways, here are some warnings noted.
|
Attachment: install.2.log A second attampt after touching the missing file. The build stops earlier, which is odd |
comment:3
I attached a copy of the complete ecl part of install.log, but this time after I touched the file $SAGE_HOME/local/include/ecl/config.h first What is odd, is that the build process aborted earlier (the log file is 379 kB, rather than the 0.7 MB ealier). I've tried downloading version 9.4.1 of ECL (as used in Sage) from the site at http://ecls.sourceforge.net/ The build of that failed. (I can't recall exactly where). I also downloaded the latest ECL (9.7.1) and that failed too. Finally I tried the latest snaphot via CVS and that would not build. I've reported this on the ECL tracker and the ECL mailing list Neither has received any response in the 4 to 5 days since I reported this. Has anyone got any idea what we might do to resolve this issue? Dave |
comment:4
A new version of ECL (9.8.1) has been released. I've created a package at http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.8.1/ecl-9.8.1.spkg It is NOT ready for review yet, as I know Maxima will break with this, due to a bug in Maxima. Dave |
comment:5
Here is an updated .spkg for ECL, which fixes the bugs which prevent ECL building on Solaris SPARC. http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.8.1/ecl-9.8.1.spkg I've created a new Maxima.spkg based on the latest 5.19.0 (trac #6999) http://sage.math.washington.edu/home/kirkby/Solaris-fixes/maxima-5.19.0/maxima-5.19.0.spkg If this updated ECL is applied, then the update to Maxima must be made, as this new ECL will not build Maxima due to a 35-year old bug in Maxima (confirmed by the Maxima developers). |
comment:7
There is a bug in ECL which is causing test failures. The bug has been identified by Juan Jose Garcia-Ripoll and I intend implementing his change and making a new .spkg within the next 18 hours (at the very most - probably less than 3 hours). Dave |
comment:8
The bug has been corrected. Some other changes were made to spkg-install too. See relevant files at http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.8.1 The ecl spkg file is: http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.8.1/ecl-9.8.1.spkg |
comment:9
There has been another update to ECL - it is now version 9.8.3 The files may be found at http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.8.3 |
comment:10
The latest version installs on an intel macbook running 10.4.11. |
comment:12
I don't have access to solaris but I tried on 32-bit scientific linux. I had problems with ecl-9.8.1 but ecl-9.8.3.spkg installed without problems. I used it to compile maxima-5.19.0 (#6699) and fricas-1.0.7. Both packages compiled and installed. The tests in various interface files: lisp.py, fricas.py, axiom.py all passed. There were issues with maxima.py but I think that is a maxima problem and not an ecl one. Adam |
comment:13
I got similar positive results with 64 bit Ubuntu 9.04. A. |
comment:14
Two small things.
Adam |
Attachment: doctest-mvngu.log full log of doctest failures due to new ecl + maxima |
comment:19
I've copied AlexGhitza's ECL and Maxima packages to my home directory: http://sage.math.washington.edu/home/mvngu/patch/ecl-9.8.4.spkg http://sage.math.washington.edu/home/mvngu/patch/maxima-5.19.1.p0.spkg These updated spkg's build fine on sage.math. But I get the following doctest failures:
A full doctest log is attached to this ticket; see |
comment:20
Oh... I didn't apply the patch at #6699. No wonder I got the doctest failures. |
Author: David Kirkby |
comment:21
Here are the results of my test on various platforms. For each of the reported platforms, I first installed ecl-9.8.4.spkg, then maxima-5.19.1.spkg, followed by applying the patch
The failures relating to graph theory modules are expected failures, as cliquer is known to fail to build under OS X 10.5.8 in 64-bit mode and under various 64-bit platforms. At least I'm glad that ecl-9.8.4.spkg and maxima-5.19.1.spkg successfully compile in 64-bit mode.
For each of the above Fedora/Red Hat platforms, the common failure is in this file:
The relevant doctests that result in the failures are:
and
A work around is to change the first doctest to:
But overall, I'm happy that the updated ECL and Maxima packages at least build on sage.math, OS X in both 32- and 64-bit modes, Solaris on t2.math, and various Fedora/Red Hat platforms. Merged the package |
Merged: Sage 4.1.2.alpha0 |
Changed reviewer from Adam Webb to Adam Webb, Alex Ghitza, Minh Van Nguyen |
comment:22
The failures of the type "gcd in t" are due to the fact that the new Maxima spkg has more commands than the old, but the old list is cached in the user's .sage directory. So one solution is to remove that file, then test again (if the file is not found, it is generated automatically, this time with the correct contents). |
comment:23
See #6883 for a follow-up to this ticket. |
I've not looked into debugging this yet, but ecl fails on Solaris. This is on a Sun Blade 2000 - not 't2' though I expect it would fail on that too.
It looks like a failure to remove the non-existant file config.h is causing this. Perhaps rm -f would be acceptable?? It's possible that rm on Solaris exits with a non-zero exit code if it can't delete a file, but on other systems this does not happen. I don't know, and have not investigated yet.
See below.
Component: packages: standard
Keywords: solaris ecl config.h
Author: David Kirkby
Reviewer: Adam Webb, Alex Ghitza, Minh Van Nguyen
Merged: Sage 4.1.2.alpha0
Issue created by migration from https://trac.sagemath.org/ticket/6564
The text was updated successfully, but these errors were encountered: