[Top][All Lists]

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

spurious install of ${prefix}/include/gps.h

From: Greg Troxel
Subject: spurious install of ${prefix}/include/gps.h
Date: Tue, 03 Aug 2021 20:03:49 -0400
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/26.3 (berkeley-unix)

Gary and I have been having an offlist discussion trying to chase down a
problem I ma having and this is a quick summary and request for others
to try a build to see if you see the problem too.  This started out as
an issue with the pkgsrc build of gpsd but I have now simplified it to
be reproducible without pkgsrc.   We've made a ton of progress and I
have arrived at a specific reproducer.

What I am seeing is that if I build such that

  environment has CPPFLAGS=-I/usr/pkg/include

  scons is passed "prefix=/usr/pkg"

then after everything is build, scons decides that -- even though this
is a build phase step, not install -- that /usr/pkg/include/gps.h is out
of date and tries to copy include/gps.h from the sources to
/usr/pkg/include/gps.h, as it would properly do during scons install.

I have tried to narrow thigsn down, and it doesn't matter that the
prefix is /usr/pkg.  I have reproduced with /usr/pkg2 and /usr/local,
including with /usr/local in CPPFLAGS and not setting prefix (letting it
be the default).

If I change -I/usr/pkg/include to to be includ (just dropping an e) the
spurious install doesn't happen.

I see the passed-in CPPFLAGS being properly added to the gcc build
lines, just as you'd expect.  So that part is working fine.

To reproduce, I recommend building as a non-root user so the copy to
prefix/include will fail.

I have seen this on NetBSD with scons 3.1.2 (and python 3.8).  Other
than this, gpsd builds and passes regression tests.

Attachment: signature.asc
Description: PGP signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]