require SAGE_CHECK to be "yes"

[jhpalmieri]

Right now, setting SAGE_CHECK to any nonempty string (e.g., "no") runs the test suite. The documentation actually says that SAGE_CHECK should be "yes" for this to happen. Fix this.

While we're at it, fix something else: in the script SAGE_ROOT/local/bin/sage-env, SAGE64 is required to be "yes", "no", or unset:

if [ "$SAGE64" != "yes" -a "$SAGE64" != "no" ]; then
    echo "The environment variable SAGE64 (=$SAGE64) must be either unset, yes or no."
    exit 1
fi

The problem is, whenever sage-env is run, output is redirected to /dev/null, so this error message isn't printed. So for example:

$ export SAGE64='maybe'
$ sage
$

Sage fails to run and is completely silent as to why. Fix this, too.

---

Also, due to a bug in sage-spkg, successful test suite runs never get logged in spkg/installed/<package-name> as they should (or is intended); this is fixed by the reviewer patch.

(Note that test suite failures cannot be logged in these files as they get deleted on non-successful builds, which [currently] includes successful builds with failing self-tests.)

---

Apply only attachment: trac_9960-scripts-SAGE_CHECK.v2.patch (to the scripts repo).

CC: @sagetrac-drkirkby @kcrisman @nexttime @qed777 @jdemeyer

Component: user interface

Author: John Palmieri, Leif Leonhardy

Reviewer: Leif Leonhardy, Ivan Andrus, David Kirkby

Merged: sage-4.7.1.alpha1

Issue created by migration from https://trac.sagemath.org/ticket/9960

[jhpalmieri]

comment:3

By the way, if you look at sage-env, it rarely produces any output, so I don't see any reason to redirect it to /dev/null. The lines

. $SAGE_ROOT/local/bin/sage-env   1>/dev/null 2>/dev/null

if [ $? -ne 0 ]; then
   echo "Error setting environment variables by running $SAGE_ROOT/local/bin/sage-env; possibly contact sage-devel (see http://groups.google.com/group/sage-devel)."
fi

from the sage-sage script don't actually seem to print the warning message; I'm not sure why. But it's clearer if it says

$ export SAGE64=dlkjdf
$ sage
The environment variable SAGE64 (=dlkjdf) must be either unset, yes or no.

[nexttime]

comment:4

I commented on this somewhere where I uploaded the "first aid" patch for spaces in SAGE_ROOT.

Redirecting the output of sage-env is really bad, and the test of $? is almost useless, since sage-env is (and must be) sourced, so it must not exit [some value].

The SAGE_CHECK issue (interpreting everything except "no" as "yes") is AFAIK a historical relict; some spkgs might still run the test suite from spkg-install if SAGE_CHECK is set and set to anything but "no".

[nexttime]

comment:5

Replacing every occurrence of exit in sage-env by return should work (then testing $? makes sense); I wonder why Sage is started even if sage-env failed though. (I think there an exit 1 is missing.)

And we should quote all occurrences of SAGE_ROOT...

[nexttime]

comment:6

Any idea why sage-build doesn't source sage-env?!?

[nexttime]

comment:7

Replying to @nexttime:

I commented on this somewhere where I uploaded the "first aid" patch for spaces in SAGE_ROOT.

#9644 (There' also a patch, including one to sage-sage.)

[jhpalmieri]

comment:8

Replying to @nexttime:

Redirecting the output of sage-env is really bad,

I agree.

and the test of $? is almost useless, since sage-env is (and must be) sourced, so it must not exit [some value].

Okay.

The SAGE_CHECK issue (interpreting everything except "no" as "yes") is AFAIK a historical relict; some spkgs might still run the test suite from spkg-install if SAGE_CHECK is set and set to anything but "no".

I ran "grep" after unpacking all of the spkg's, and found SAGE_CHECK used in cliquer and mpir. In both cases, it checks whether it's "yes". (In the case of cliquer, this is done in spkg-install, but there is no spkg-check. So it does the right thing -- running the test suite once if SAGE_CHECK is "yes", just not in the right way. In the case of mpir, if SAGE_CHECK is yes, then the test suite gets run twice, once because of the main sage-spkg script, and once more in mpir's spkg-install script. I thought there was a ticket for this, but I can't find it right now.)

Replacing every occurrence of exit in sage-env by return should work (then testing $? makes sense); I wonder why Sage is started even if sage-env failed though. (I think there an exit 1 is missing.)

When do you see Sage starting if sage-env fails?

I'll also take a look at #9644. If I can give that a positive review, I'll change the patch here to make it depend on #9644. (I may do that anyway...) As far as changing sage-build to source sage-env, could that break things? The other changes here are pretty inocuous.

[jhpalmieri]

comment:9

Ah, the mpir issue is covered by #9522 and #8664.

[nexttime]

comment:10

Replying to @jhpalmieri:

Replying to @nexttime:

The SAGE_CHECK issue (interpreting everything except "no" as "yes") is AFAIK a historical relict; some spkgs might still run the test suite from spkg-install if SAGE_CHECK is set and set to anything but "no".

I ran "grep" after unpacking all of the spkg's, and found SAGE_CHECK used in cliquer and mpir. In both cases, it checks whether it's "yes".

Ok. Maybe we already corrected the others.

Replacing every occurrence of exit in sage-env by return should work (then testing $? makes sense); I wonder why Sage is started even if sage-env failed though. (I think there an exit 1 is missing.)

When do you see Sage starting if sage-env fails?

My bad. I think this happened after I had replaced exit by return in sage-env (screen buffer):

Error setting environment variables by running /home/leif/Sage/sage-4.6.alpha1-final/local/bin/sage-env; possibly contact sage-devel (see http://groups.google.com/group/sage-devel).
----------------------------------------------------------------------
| Sage Version 4.6.alpha1, Release Date: 2010-09-18                  |
| Type notebook() for the GUI, and license() for information.        |
----------------------------------------------------------------------
**********************************************************************
*                                                                    *
* Warning: this is a prerelease version, and it may be unstable.     *
*                                                                    *
**********************************************************************
sage:

(But the same happens if source itself fails, e.g. because sage-env isn't found, likewise when SAGE_ROOT contains spaces.)

But if the error message ever gets printed, Sage shouldn't start. We currently have:

. $SAGE_ROOT/local/bin/sage-env   1>/dev/null 2>/dev/null

if [ $? -ne 0 ]; then
   echo "Error setting environment variables by running $SAGE_ROOT/local/bin/sage-env; possibly cont
act sage-devel (see http://groups.google.com/group/sage-devel)."
fi

...

sage() {
    sage_setup
    sage-ipython "$@" -i
}

if [ $# -eq 0 ]; then
    sage
    exit $?
fi

As far as changing sage-build to source sage-env, could that break things? The other changes here are pretty inocuous.

I guess that's just an old flaw; calling it is of little use (will only check some things, but doesn't change the environment at all).

So give it a try and see what happens... ;-)

(It shouldn't^TM break anything.)

[nexttime]

comment:11

Since sage-build is (or should be) called from sage-sage, we can either source it again or completely omit the (second) call.