"make" should run Sage once

[jdemeyer]

In a multi-user environment, the user compiling Sage must run it at least once to run sage-location and generate .pyc files.

The proposed fix is: in the default make rule, run Sage when local/bin/sage-started.txt does not exist and create that file in sage-location. Also run Sage after upgrading.

This patch also changes the error message when a spkg fails a build or test. Example error message:

Error: Configuring PARI with readline and GMP kernel failed.

real    0m0.100s
user    0m0.012s
sys     0m0.012s
************************************************************************
Error installing package pari-2.4.3.alpha.p7
************************************************************************
Please email sage-devel (http://groups.google.com/group/sage-devel)
explaining the problem and including the relevant part of the log file
  /usr/local/src/sage-4.7.2.rc0/spkg/logs/pari-2.4.3.alpha.p7.log
Describe your computer, operating system, etc.
If you want to try to fix the problem yourself, *don't* just cd to
/usr/local/src/sage-4.7.2.rc0/spkg/build/pari-2.4.3.alpha.p7 and type 'make' or whatever is appropriate.
Instead, the following commands setup all environment variables
correctly and load a subshell for you to debug the error:
  (cd '/usr/local/src/sage-4.7.2.rc0/spkg/build/pari-2.4.3.alpha.p7' && '/usr/local/src/sage-4.7.2.rc0/sage' -sh)
When you are done debugging, you can type "exit" to leave the subshell.
************************************************************************
Error: Failed to install package 'pari'.

Apply:

1. attachment: 11926.patch to SAGE_ROOT
2. attachment: 11926_sage_starts.patch, attachment: trac_11926-error-msg.patch, attachment: 11926-error-msg-review.patch to SCRIPTS (local/bin)
3. attachment: 11926_sage.patch and attachment: 11926_doc.patch to the Sage library.

CC: @pcpa

Component: build

Keywords: Makefile build sage-starts

Author: Jeroen Demeyer

Reviewer: John Palmieri, Leif Leonhardy

Merged: sage-4.8.alpha0

Issue created by migration from https://trac.sagemath.org/ticket/11926

[jdemeyer]

Description changed:

--- 
+++ 
@@ -1,3 +1,3 @@
 In a multi-user environment, the user compiling Sage must run it at least once to generate the `sage-current-location` file, generate `.pyc` files, update package-config files.
 
-The proposed fix is: in the top-level Makefile, if `local/lib/sage-current-location.txt` does not exist, do a dummy run of Sage.
+The proposed fix is: in `spkg/install`, run Sage.

[jhpalmieri]

comment:3

I think I suggested this sort of thing once, and it was criticized. I don't remember the critiques. Anyway, I like the idea. In case some system administrator is building Sage and doesn't want a .sage directory created for them, should we set DOTSAGE to something temporary first?

By the way, running Sage as part of the build process should also take care of some of the issues at #5155 (not all of them, unfortunately).

[jdemeyer]

comment:4

Replying to @jhpalmieri:

I think I suggested this sort of thing once, and it was criticized. I don't remember the critiques. Anyway, I like the idea.

In case some system administrator is building Sage and doesn't want a .sage directory created for them, should we set DOTSAGE to something temporary first?

Good idea. Make a temporary DOTSAGE, run Sage, remove the temporary DOTSAGE

By the way, running Sage as part of the build process should also take care of some of the issues at #5155 (not all of them, unfortunately).

Thanks for the pointer. Also #11887 is related.

[jdemeyer]

Work Issues: temporary DOTSAGE

[nexttime]

comment:5

I don't like the idea of running sage after make build since it certainly doesn't belong to that target, but if at all, it shouldn't be run from spkg/install but the Makefile, preferably in the default rule, not build. Or run sage-location from there if make did anything, but we should in principle run it from elsewhere.

Someone who is going to install Sage system-wide should run make ptestlong or similiar anyway (which also runs sage), just like one runs make check before running make install.

Also note that sage (or sage-location) should be run (at least) after it has been moved to its final installation location, which is likely to happen after having built Sage.

[jdemeyer]

comment:6

Replying to @nexttime:

I don't like the idea of running sage after make build since it certainly doesn't belong to that target

I think it does. Currently, after make build (or simply make), Sage does not work for a different user. I think make build should result in a proper Sage installation, which means that it should work for every user.

it shouldn't be run from spkg/install but the Makefile

Why? I like spkg/install better because it allows us to run Sage only if something was actually installed. You don't want to run Sage everytime somebody does make foo for various values of "foo". Also, IIRC, we run spkg/install on upgrading but not a top-level make all or make build.

Or run sage-location from there if make did anything, but we should in principle run it from elsewhere.

Running sage-location is certainly not sufficient to generate all files needed to run Sage.

Someone who is going to install Sage system-wide should run make ptestlong

Let the sysadmin decide whether they want to run make ptestlong or not, I want to give him the option of NOT doing it.

Also note that sage (or sage-location) should be run (at least) after it has been moved to its final installation location, which is likely to happen after having built Sage.

I would think most people run Sage from the location where they installed it, or equivalently, install Sage in the location they want to run it from. In any case, this does not invalidate the reasons for this ticket.