Peter Jeremy wrote:
> On 2009-Aug-02 12:24:26 -0700, William Stein <wst...@gmail.com> wrote:
>>> 1) It uses 'set -e' to exit the spkg-install if any error occurs.
>> +1 we should do that for all spkg-install's
> 
> Agreed (and there are other shell-scripts that probably should also
> have 'set -e').  But this is a very intrusive change since it touches
> at least one file in virtually every spkg.
> 
> It's also non-trivial:  The spkg-install files are full of constructs
> like:
>   sage_fortran -c -fPIC *.f
>   if [ $? -ne 0 ]; then
>     echo "Error compiling blas."
>     exit 1
>   fi
> In order to add 'set -e', this needs to be changed to directly test the
> exit code rather than using $? - otherwise the script will immediately
> exit when the command fails.  The revised code should look like:
>   if ! sage_fortran -c -fPIC *.f; then
>     echo "Error compiling blas."
>     exit 1
>   fi
> 
> In addition to the above, there is a non-trivial amount of shell code
> embedded in python scripts that also needs improved error checking.
> 
> Assuming that the change should be implemented, what approach should
> be used:
> 1) Big bang: Someone goes through all the skpg files and fixes them
>    in one go with no other changes as part of a major release.
> 2) If you touch any part of an spkg, you must add 'set -e' and fixup
>    any subsequent fallout in all scripts in that spkg.
> 3) If you touch a script, you must fixup that script.
> 4) Fix it if you feel like it.
> 

I think (2). As you note, it does need some care when applied. If it is 
applied where it should not, it will not give a useful error message, 
but it will at least exit. One could reasonably say that is better than 
carrying on, when something is broken.


--~--~---------~--~----~------------~-------~--~----~
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---

Reply via email to