document more environment variables

[John Palmieri]

In ticket #8263, we missed these:

• SAGE_DOC_JSMATH - should force docs to be built using jsMath

• SAGE_DOCBUILD_OPTS - passed as an argument to sage --docbuild in the main Sage makefile.

[Leif Leonhardy]

I'm quite sure there are more missing... ;-)

[Leif Leonhardy]

TODO:

• Add PARI_EXTRA_OPTS, which is passed to PARI's Configure when building Sage / installing the PARI spkg.
• Add CPPFLAGS, LDFLAGS, CXXFLAG64?, LDFLAG64? LD?
• Document that (e.g.) export CFLAGS="" has not the same effect as unset CFLAGS.
• Add note on not putting a space between -j and NUM in the description of MAKE, because doing so has a "completely" different meaning...
• Add warning to (setting) SAGE_CHECK, since this currently breaks most builds. This will (or should) be fixed in future releases though; there's afaik a ticket adding more options to the value of this variable...
• Add a note on / reference to sage-env.

Also, the description of some variables is incomplete, not current (in its use in all spkgs, e.g. SAGE_DEBUG), misleading, or even wrong (SAGE_TIMEOUT*).

I think the various variations of SAGE_T[E]MP[[_]DIR] are missing. (I wonder where e.g. TMP and TMPDIR are / may be used instead, or as a fall-back / default.)

SHELL might be relevant, too.

The *ITER* variables for (parallel) doctesting are missing.

I'm currently not sure if we have a similar section in the Developer's Guide, but some variables like SAGE_ROOT and SAGE_LOCAL should be documented (there?), too.

I'm still pretty sure even more variables are missing... ;-)

[Karl-Dieter Crisman]

SAGE_DOC_JSMATH

SAGE_DOC_MATHJAX is now in good shape.

• SAGE_DOCBUILD_OPTS

if it still exists

• Add PARI_EXTRA_OPTS, which is passed to PARI's Configure when building Sage / installing the PARI spkg.
• Add CPPFLAGS, LDFLAGS, CXXFLAG64?, LDFLAG64? LD?
• Document that (e.g.) export CFLAGS="" has not the same effect as unset CFLAGS.
• Add note on not putting a space between -j and NUM in the description of MAKE, because doing so has a "completely" different meaning...
• Add warning to (setting) SAGE_CHECK, since this currently breaks most builds. This will (or should) be fixed in future releases though; there's afaik a ticket adding more options to the value of this variable...

This is basically fine now

• Add a note on / reference to sage-env.

Also, the description of some variables is incomplete, not current (in its use in all spkgs, e.g. SAGE_DEBUG), misleading, or even wrong (SAGE_TIMEOUT*). I think the various variations of SAGE_T[E]MP[[_]DIR] are missing. (I wonder where e.g. TMP and TMPDIR are / may be used instead, or as a fall-back / default.) SHELL might be relevant, too. The *ITER* variables for (parallel) doctesting are missing. I'm currently not sure if we have a similar section in the Developer's Guide, but some variables like SAGE_ROOT and SAGE_LOCAL should be documented (there?), too.

There is a ticket for the latter to be better, and the former is now in good shape.

I'm still pretty sure even more variables are missing... ;-)

Indubitably.

[John Palmieri]

Replying to kcrisman:

• SAGE_DOCBUILD_OPTS

if it still exists

It is actually used for docbuilding in the top-level Makefile. So it ought to be documented.

[John Palmieri]

New commits:

​13513fa

Trac 9440: Document SAGE_DOCBUILD_OPTS

[Karl-Dieter Crisman]

Nice! Though I misled you by the wiki formatting. I don't know the status of these, and at least some definitely aren't documented.

• Add PARI_EXTRA_OPTS, which is passed to PARI's Configure when building Sage / installing the PARI spkg.

Needed?

• Add CPPFLAGS, LDFLAGS, CXXFLAG64?, LDFLAG64? LD?

Needed?

• Document that (e.g.) export CFLAGS="" has not the same effect as unset CFLAGS.

Needed?

• Add note on not putting a space between -j and NUM in the description of MAKE, because doing so has a "completely" different meaning...

Needed?

• Add a note on / reference to sage-env.

Needed?

Also, the description of some variables is incomplete, not current (in its use in all spkgs, e.g. SAGE_DEBUG), misleading, or even wrong (SAGE_TIMEOUT*).

Needed?

I think the various variations of SAGE_T[E]MP[[_]DIR] are missing. (I wonder where e.g. TMP and TMPDIR are / may be used instead, or as a fall-back / default.)

Needed?

SHELL might be relevant, too.

Needed?

The *ITER* variables for (parallel) doctesting are missing.

Needed?

[Karl-Dieter Crisman]

Indeed, the output of

sage.env.[tab]

is also worth considering.

[git]

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

​b7182fb	Trac 9440: document some more environment variables.
​79280e1	Trac 9440: document SAGE_TEST_*_ITER

[John Palmieri]

Replying to kcrisman:

Nice! Though I misled you by the wiki formatting. I don't know the status of these, and at least some definitely aren't documented.

• Add PARI_EXTRA_OPTS, which is passed to PARI's Configure when building Sage / installing the PARI spkg.

Needed?

Doesn't seem to be used, but some other variables are used in Pari's spkg-install file. I've documented those (just copying from spkg-install, basically).

• Add CPPFLAGS, LDFLAGS, CXXFLAG64?, LDFLAG64? LD?

Needed?

I added a bit about them.

• Document that (e.g.) export CFLAGS="" has not the same effect as unset CFLAGS.

Needed?

Same here.

• Add note on not putting a space between -j and NUM in the description of MAKE, because doing so has a "completely" different meaning...

Needed?

I don't know, this isn't Sage specific. I think we can skip it.

• Add a note on / reference to sage-env.

Needed?

Done.

Also, the description of some variables is incomplete, not current (in its use in all spkgs, e.g. SAGE_DEBUG), misleading, or even wrong (SAGE_TIMEOUT*).

Needed?

I noticed that the documented value for SAGE_TIMEOUT was wrong. Beyond that, I don't know what Leif meant by this. If he wants to expand on it at some point, he can open another ticket and clarify.

I think the various variations of SAGE_T[E]MP[[_]DIR] are missing. (I wonder where e.g. TMP and TMPDIR are / may be used instead, or as a fall-back / default.)

Needed?

I don't think there are such variables anymore. There is a variable SAGE_TMP used while running Sage, but it's not user-customizable and it's not an environment variable. So I don't think we need to document it.

SHELL might be relevant, too.

Needed?

Why? Let's skip it. What does it have to do with Sage?

The *ITER* variables for (parallel) doctesting are missing.

Needed?

I added something about them.

[git]

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

​dbea30c

Trac 9440: change date in source.rst

[John Palmieri]

Oh, for what it's worth, I set SAGE_TUNE_PARI to yes and rebuilt pari. Without this set, it took 4 minutes. With it set, it took 2 hours. So the comment about being time-consuming is accurate.