pkgsrc-Changes archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: CVS commit: pkgsrc/pkgtools/pkg_rolling-replace

On 02/01/10 13:32, Greg Troxel wrote:

   >  I also don't understand the -j option.  While I want pkg_rr to "just
   >  work", I also leave PKG_DEVELOPER=YES and MAKE_JOBS=2 in mk.conf so I
   >  can fix/notice underlying package problems.

   But I don't want MAKE_JOBS=2 being always active. E.g. when updating a
   package (let's name it math/pari) and want to see the compile errors.

So you want to set MAKE_JOBS=2 when doing testing of packages, but turn
it off when doing rebuilds?

No, I test with (to ensure, MAKE_JOBS are safe) and without MAKE_JOBS=n (mostly when updating and errors came up). But I don't choose n=4*CPU (which is the best value for optimal load) during tests. For rebuild my system (including X11, TeX, XFCE4,, ...) I use a much larger value than during developing (usually 4*CPU - read -j8 on Dual-Core CPU, -j32 on 2xQuad-Core etc.).

I wonder if having pkg_rr define some sort of variable to indicate that
we're in a bulk update or pkg_rr mode would help, and then you can put
in mk.conf the equivalent of:

if in_pkg_rr
   blah blah

and people can set whatever they want conditionally.

Good idea \o/ /o\

   >  Then set USE_DESTDIR=no ...

   I've seen that pkg_rr doesn't transports environment variables to
   ${MAKE}. And there's a difference for me between developing on pkgsrc
   and using pkgsrc. During developing, I'm present and act on
   errors. During using (e.g. update X11&  Co.), I'm not present and
   simply want it working.

I can certainly see why you have that as a goal.  I wouldn't object to a
generic "set this in the make environment" option, but I am wondering if
my suggestion of a pkg_rr-specific variable always defined would make
that unnecessary.

No, you're suggestion is much better. I'll try to change it today - but I want to go further with my current work on math/p5-Math-Pari first.


Home | Main Index | Thread Index | Old Index