Update Valgrind to 3.10.0

[jpflori]

New version is out.
Use git branch and tarball at:

• http://sage.sagedev.org/home/vbraun/upstream/valgrind-3.10.0.tar.bz2

CC: @vbraun @gvol

Component: packages: optional

Keywords: spkg valgrind

Author: Jean-Pierre Flori, Volker Braun

Branch/Commit: e150b39

Reviewer: Jeroen Demeyer

Issue created by migration from https://trac.sagemath.org/ticket/15586

[jpflori]

Changed keywords from none to spkg valgrind

[jpflori]

Branch: u/jpflori/ticket/15585

[jpflori]

comment:1

Sole non-trivial modification was to remove a patch now upstream.
Whatsoever, we should still take care of #14097.

[jpflori]

New commits:

9d5a3bf

Update Valgrind to 3.9.0.

[jpflori]

Commit: 9d5a3bf

[jpflori]

Changed branch from u/jpflori/ticket/15585 to u/jpflori/ticket/15586

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

556c5f4

Update Valgrind to 3.9.0.

[sagetrac-git]

Changed commit from 9d5a3bf to 556c5f4

[vbraun]

comment:5

tarball is go-r

vbraun@boxen:~$ ll /home/jpflori/upstream/valgrind-3.9.0.tar.bz2
-rw------- 1 jpflori jpflori 10003156 2013-12-25 08:42 /home/jpflori/upstream/valgrind-3.9.0.tar.bz2

[jpflori]

comment:6

Fixed.

[jdemeyer]

comment:10

I dislike

+if [ "$UNAME" = "Linux" ]; then
+ echo "Good - Valgrind works on Linux"
+ if [ "`uname -p`" = "ia64" ]; then
+ echo >&2 "But it does not work on Itanium"
+ exit 1
+ fi
+else
+ if [ "$UNAME" = "Darwin" ] && [ -z "`uname -p | grep PPC`" ] && [ `uname -r | grep -Eo '^[0-9]+'` -ge 9 ];
+ then
+ echo "Good - Valgrind works on x86 and AMD64 Darwin 9.x or 10.x"
+ else
+ echo >&2 "Sorry, Valgrind only works on x86,AMD64,PPC32,PPC64,ARM Linux"
+ echo >&2 "and x86,AMD64 on Darwin 9.x/10.x"\
+ "(Mac OS X 10.5.x or later)"
+ exit 1
+ fi
+fi

If it doesn't work, fine, just let the build fail (one would hope that upstream gives a reasonable error message in this case). Too often, these kind of checks prevent building on systems which are perfectly fine.

[jdemeyer]

comment:11

Also, the suppressions are not patches so I see no reason to put them in the patches directory.

[jdemeyer]

Reviewer: Jeroen Demeyer

[jpflori]

comment:13

The good/bad stuff is historical.

As far as non patch stuff in the patches directory, I had the feeling some other packages did that.

And my first concern here is that I tested the package a couple of days ago and:

• it does not build at first, stupid glibc version restriction as usual with valgrind...
• and it does not work properly, or at least did not for me: at least Sage now launches (not as reported in Fix broken 'sage -valgrind' interaction with ipython #14097 which I could reproduce when the ticket was active), but I cannot exit it...

[jpflori]

comment:14

(I tested it on ppc64 and x86_64.)

[vbraun]

Changed author from Jean-Pierre Flori to Jean-Pierre Flori, Volker Braun

[vbraun]

Changed branch from u/jpflori/ticket/15586 to u/vbraun/valgrind_3_10

[vbraun]

Changed commit from 556c5f4 to bed4a70

[vbraun]

comment:17

I'm moving the suppression files to $SAGE_EXTCODE/valgrind/sage.supp, that way we can even use the system valgrind (which is also more likely to work with your version of glibc).

---

New commits:

7b1989d	Update Valgrind to 3.9.0.
bed4a70	Update Valgrind to 3.10.0

[sagetrac-git]

Changed commit from bed4a70 to 2fb8eb3

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

2fb8eb3

Move suppression files to extcode

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

8d962d2

Add PyAlloc suppressions

[sagetrac-git]

Changed commit from 2fb8eb3 to 8d962d2

[vbraun]

comment:20

Works and is actually useful (doesn't spew spurious pyalloc warnings all the time)

[jdemeyer]

comment:21

Replying to @jpflori:

The good/bad stuff is historical.

Perhaps, but I still think it should be removed.

[sagetrac-git]

Changed commit from 8d962d2 to e150b39

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

e150b39

Remove unnecessary checks

[vbraun]

comment:23

Fixed.

[jpflori]

comment:24

Volker: One naive question: is the sage -valgind stuff actually working for you? last time I tested (see above), everything was actually broken except for Sage starting.)

[vbraun]

comment:25

Yes, works. Fixed that, too :-)

[jpflori]

comment:26

It seems there is an issue with the new location of the supp files and where the sage-valgrind script looks for.

[vbraun]

comment:27

Whats the issue? You need to run "make" to install it to SAGE_EXTCODE.

[jpflori]

comment:28

Oh my bad....
Then everything is OK.

[jpflori]

comment:29

I'm still getting thousands of errors when starting up. Is that expected ?

[vbraun]

comment:30

Did you recompile python, if not then the python suppression file is not installed. The patch doesn't update the python version so it is not recompiled automatically.

[jpflori]

comment:31

Nope I did not.

[vbraun]

Changed branch from u/vbraun/valgrind_3_10 to e150b39