[csw-maintainers] surprising set of packages when build spans 2 days

[pfelecan]

I started a build on February 25th which ended on 26th:

    nohup mgar spotless platforms < /dev/null > ~/logs/texlive 2>&1 &

The resulting packages are in ~/staging/build-25.Feb.2013

Here are the first lines of the content as shown by
    ls -1 ~/staging/build-25.Feb.2013

    libkpathsea6-20120701,REV=2013.02.25-SunOS5.10-sparc-CSW.pkg.gz
    libkpathsea6-20120701,REV=2013.02.26-SunOS5.10-i386-CSW.pkg.gz
    libptexenc1-20120701,REV=2013.02.25-SunOS5.10-sparc-CSW.pkg.gz
    libptexenc1-20120701,REV=2013.02.26-SunOS5.10-i386-CSW.pkg.gz
    pdfjam_stub-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    pdfjam_stub-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    tetex_stub-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    tetex_stub-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    
tex_pdftricks_old_stub-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    
tex_pdftricks_old_stub-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    tex_pdftricks_stub-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    tex_pdftricks_stub-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    texlive-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    texlive-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    texlive_base-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    texlive_base-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    
texlive_bibtex_extra-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    
texlive_bibtex_extra-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz
    texlive_binaries-20120701,REV=2013.02.25-SunOS5.10-sparc-CSW.pkg.gz
    texlive_binaries-20120701,REV=2013.02.26-SunOS5.10-i386-CSW.pkg.gz
    texlive_common-20120701,REV=2013.02.25-SunOS5.10-all-CSW.pkg.gz
    texlive_common-20120701,REV=2013.02.26-SunOS5.10-all-CSW.pkg.gz

The sparc specific packages were generated before midnight, the
i386 ones after midnight.

As we can see, all the architecture neutral packages, infix "all",
are duplicated. This is not only a waste of space but also of
time: it means that for each platform they are generated
even though once is enough.

Can this be optimized by generating only once the "all" packages
when using the "platforms" target?

I don't know which is the effect of this on the packages upload
but I'll test it very soon.