pkgsrc-Users archive

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

Re: pkgsrc-2017Q3 installation on 6.1.5 i386



"Nick B. (lists)" <nick.netbsd%nowindows.net@localhost> writes:

> I had a working pkgsrc-2016Q3 on my system. I decided to upgrade this
> to pkgsrc-2017Q3.
>
> Various packages fail to build. Can I define a later version of gcc
> (gcc48) in /usr/pkg/etc/mk.conf so it gets built and is then used to
> build everything else from pkgsrc? From reading the pkgsrc guide it
> appears that I need to add
> GCC_REQD=4.8
> to mk.conf to force this. If this isn't the best way to get up and
> running I'd be grateful for some guidance.

The right way is to finish the compiler selection logic we've been
discussing :-)  See https://wiki.netbsd.org/pkgsrc/gcc/ for some
discussion of where we are.

But seriously....

What I would do is one of:

1) upgrade to netbsd-7.  This has gcc 4.8, and most things are ok.  If
you can do this, you take a lot of pain off the table.  Part of the
problem is that most people who work on NetBSD have moved beyond -6 on
most systems.

2) Do as you say, but I'd choose 4.9, since you have to build a new
version anyway, and at least firefox needs 4.9, so if you want that and
you used 4.8, you'd end up with both.

   In general, you should be building all C++ stuff with the same
   compiler.  GCC_REQD=4.9 probably does not play well with building gcc
   4.9.  Assuming you are using pkg_rolling-replace, I would:

   cd /var/db/pkg && pkg_admin set rebuild=YES *
   # or, if you are more adventurous, only do this for packages that use
   # c++.  Or skip it and look for trouble later.

   without GCC_REQD=4.9 in mk.conf
   [perhaps set the gcc-inplace-math option to avoid two dependencies
   with a different compiler than the rest]
   build/install the gcc 4.9 package

   put GCC_REQD=4.9 in mk.conf

   start "pkg_rolling-replace -uvk", or leave off the -k and think about
   each problem

3) live dangerously by just "make GCC_REQD=4.9 package replace" for each
package that fails, but then you end up with mixed 4.5/4.9 for C++.  But
if it's only a few usual-suspect packages like harfbuzz that get used by
only pango, you can perhaps get away with this, even though it's in
general unsound.

Attachment: signature.asc
Description: PGP signature



Home | Main Index | Thread Index | Old Index