Memory leaks with LP Solvers

[ncohen]

As reported on sage-support, there is(was) something dead wrong with the LP classes : they did not free some of the memory they used.

This patch solves it !

Nathann

Reported by Peter Mueller on  sage-support.

---

Apply: trac_11917.patch

(Based on Sage 4.7.2.alpha3 / #11588.)

[zimmerma]

the patch fails to apply to 4.7.1:

----------------------------------------------------------------------
| Sage Version 4.7.1, Release Date: 2011-08-11                       |
| Type notebook() for the GUI, and license() for information.        |
----------------------------------------------------------------------
Loading Sage library. Current Mercurial branch is: 11917
sage: hg_sage.import_patch("/tmp/trac_11917.patch")
cd "/usr/local/sage-4.7.1/sage/devel/sage" && hg status
cd "/usr/local/sage-4.7.1/sage/devel/sage" && hg status
cd "/usr/local/sage-4.7.1/sage/devel/sage" && hg import   "/tmp/trac_11917.patch"
applying /tmp/trac_11917.patch
patching file sage/numerical/backends/coin_backend.pxd
Hunk #1 FAILED at 97
1 out of 1 hunks FAILED -- saving rejects to file sage/numerical/backends/coin_backend.pxd.rej
patching file sage/numerical/backends/coin_backend.pyx
Hunk #3 FAILED at 986
1 out of 3 hunks FAILED -- saving rejects to file sage/numerical/backends/coin_backend.pyx.rej
abort: patch failed to apply

Paul

[leif]

Replying to zimmerma:

the patch fails to apply to 4.7.1:

That's most probably due to #11588, merged into Sage 4.7.2.alpha3.

[zimmerma]

That's most probably due to #11588, merged into Sage 4.7.2.alpha3.

if so it should be clearly indicated to which version the patch applies.

Paul

[leif]

Patch also applies clean to Sage 4.7.2.alpha4.

IMHO all patches should in general be based on the latest devel release (or at least still apply to it); an attachment comment stating on which version a patch is based is of course not bad, especially as time goes by.

I really don't understand why the patchbot still uses 4.7.1; as such, it is almost useless. Worse, it reports failures for patches that apply (and work) well. (And often hangs such that the HTTP connections for trac pages are kept open...)

[leif]

Passes all (long) tests in sage/numerical/, and Peter Mueller's example (at first glance) doesn't eat up memory as it did before.

Haven't looked at the code though.

[pmueller]

The patch installed fine in sage-4.7.2alpha, and passed all the tests in sage/numerical/

I ran my original program again which goes through thousands of different small ilp problems. The memory usage was basically constant, so the memory leak apparently is fixed.

As this patch just fixed a bug and didn't add new features, I guess it is save to close the ticket.

• Peter Mueller

[ncohen]

Thaaaaaaaaaaaaaaaaaanks ! :-)

Nathann

[zimmerma]

why did the milestone change? Is 4.7.2 already out? I don't see it on sagemath.org.

Paul

[jdemeyer]

Replying to zimmerma:

Is 4.7.2 already out?

Not yet, but it should be almost out. A release candidate is in preparation.

[leif]

Replying to jdemeyer:

Replying to zimmerma:

Is 4.7.2 already out?

Not yet, but it should be almost out. A release candidate is in preparation.

So a small patch only fixing a memory leak isn't worth getting into the final?

[zimmerma]

I wonder why you already work on sage-4.7.3, since 4.7.2 is not yet available (even in source form) on sagemath.org...

Paul

[jdemeyer]

Milestone sage-4.7.3 deleted