Remove debugging output from Python command line

[jpflori]

The extra debugging output "[number of refs]" before each command line prompt breaks the pynad build. the problem is that the test used by the AX_PYTHON_DEVEL autoconf macro to check for distutils fails when Python is built with pydebug.

Also, breaks the Sage test suite.

I'm just removing this extra output. Its not that useful, really, since object counts in Sage are always pretty large and widely fluctuating with stuff like coercion running in the background. It also doesn't offer any extra info over gc.get_count().

Install http://www.stp.dias.ie/~vbraun/Sage/spkg/python-2.7.3.p5.spkg

CC: @burcin @vbraun @simon-king-jena

Component: packages: standard

Keywords: python spkg debug

Author: Volker Braun

Reviewer: Jean-Pierre Flori

Merged: sage-5.7.beta4

Issue created by migration from https://trac.sagemath.org/ticket/13867

[vbraun]

comment:1

I'm thinking we should just patch Python to not print out extra crap. Otherwise we have basically no hope of running the Sage doctests, no?

[jpflori]

comment:2

Good point.
And that should be even easier.

[jpflori]

comment:3

In fact I do not really agree, especially now that I've checked that the crappy additional Python output does not pertub the doctest framework (surely because its spit on stderr and not stdout).

[vbraun]

comment:4

Surely the sage pexpect interface to itself will not work. Or various checks in sage.tests.cmdline that check that stderr is empty.

[vbraun]

Author: Volker Braun

[vbraun]

comment:5

I'm taking over this ticket to patch up python

[vbraun]

Changed keywords from pynac spkg debug to python spkg debug

[simon-king-jena]

comment:7

Just to be on the safe side: To use this one with SAGE_DEBUG=yes on top of an existing Sage install, one needs to install this new python spkg, then (re-)install the new cython spkg from #13832, and then sage -ba, right?

[vbraun]

comment:8

to get a working Sage you also need a patch to the Sage library (will post here soon) and a fixed Singular spkg... stay tuned ;-)

[vbraun]

comment:9

Patch to the Sage library is at #13868

[simon-king-jena]

comment:10

We now have an abundance of tickets needed to make SAGE_DEBUG work. Shouldn't there be a meta-ticket, listing all relevant single-purpose tickets (#13876, #13867, #13868, #13832, #13864, and what else)?

[vbraun]

comment:11

#13864 is now the meta-ticket.

[simon-king-jena]

comment:12

Replying to @simon-king-jena:

We now have an abundance of tickets needed to make SAGE_DEBUG work. Shouldn't there be a meta-ticket, listing all relevant single-purpose tickets (#13876, #13867, #13868, #13832, #13864, and what else)?

FWIW: I created a metaticket at #13877.

[simon-king-jena]

comment:13

Replying to @simon-king-jena:

Replying to @simon-king-jena:

We now have an abundance of tickets needed to make SAGE_DEBUG work. Shouldn't there be a meta-ticket, listing all relevant single-purpose tickets (#13876, #13867, #13868, #13832, #13864, and what else)?

FWIW: I created a metaticket at #13877.

OK, forget about that, Volker was faster with his change to #13864.

[jdemeyer]

comment:14

In spkg-install, are you sure you don't mean CFLAGS and/or CXXFLAGS instead CPPFLAGS? Because CPPFLAGS normally means "C preprocessor flags".

[jdemeyer]

Dependencies: #13864

[jdemeyer]

Changed dependencies from #13864 to none

[vbraun]

comment:17

The spkg-install already used CPPFLAGS before I edited it, so I stuck with it. It does get picked up on the gcc command line, fwiw.

[jdemeyer]

comment:18

Yes, it does use CPPFLAGS for its intended purpose. The -I option is related to preprocessing and belongs in $CPPFLAGS. Flags like -g or -O0 do not.

Note that adding -I$SAGE_LOCAL/include to CPPFLAGS is actually unneeded since that path gets picked up automatically because of the $CPATH environment variable. So you might as well remove

CPPFLAGS="-I$SAGE_LOCAL/include $CPPFLAGS"
export CPPFLAGS

[jdemeyer]

comment:19

Replying to @vbraun:

The spkg-install already used CPPFLAGS before I edited it

It also uses CFLAGS by the way:

export CFLAGS="$CFLAGS `testcflags.sh -fwrapv`"

[jpflori]

comment:20

I'm ok with the solution Volker proposes, we don't need that "extra" output, but I still think that Pynac should be able to deal with a vanilla Python debug build.
So I'm just posting that here so that Burcin can take notice (I'll post on pynac-devel as well).

[vbraun]

Attachment: python-p5.patch.gz

diff for review only

[vbraun]

comment:21

I've removed the superfluous CPPFLAGS. Also --with-pydebug implies -O0 -g so we don't have to set it by hand.

[simon-king-jena]

comment:23

I use the package successfully with SAGE_DEBUG=yes.

Some criticism: SPKG.txt contains a section "Patches", but the new patch disable_print_refs_debug.patch is not mentioned.

The content of environment variables is sometimes tested like

if ["x$SAGE_DEBUG"="xyes"]; then

if I am not mistaken, but here it is

if ["$SAGE_DEBUG"="yes"]; then

Is that a problem?

[vbraun]

comment:24

I've added a line documenting the disable_print_refs_debug.patch

The xyes thing is a historical artifact.

[jpflori]

Reviewer: Jean-Pierre Flori

[jpflori]

comment:25

Looks good.

If you really wants me to rant, I'd say:

• first, there is a useless line at the end of the new patch for some emacs generated tmp file.
• or even less problematic, you could remove the historical artifacts from spkg-install while you're at it.

[jdemeyer]

comment:26

Server down :-(

[jdemeyer]

Merged: sage-5.7.beta4