Closure size WIP

[avnik]

Fix qtile, fvwm and subversion

[mention-bot]

By analyzing the blame information on this pull request, we identified @kamilchm, @iyzsong and @edolstra to be potential reviewers

[avnik]

I am and @lethalman decided to postpone this PR before master will be merged to closure-size, and I rebase stuff on it.

[vcunat]

Ping. I pushed a tested version of closure-size with new staging in it. (All packages I use do build.)

[vcunat]

And, in the meantime subversion started to build again, so it's possible some parts of this PR won't be needed anymore (and/or will conflict).

[avnik]

On Wed, Jan 20, 2016 at 10:51:22AM -0800, Vladimír Čunát wrote:

And, in the meantime subversion started to build again,
so it's possible some parts of this PR won't be needed
anymore (and/or will conflict).

I am not sure, if it was common problem, or it unbuildable on
my localhost only. Need to check it separately.

[vcunat]

No, at the time of filing this PR I did experience svn build problems, but I do not anymore (for the same attribute).

[avnik]

On Wed, Jan 20, 2016 at 11:54:37AM -0800, Vladimír Čunát wrote:

No, at the time of filing this PR I did experience svn build
problems.

Can you tell what problems, and how you fix them?

[vcunat]

It didn't build, but I don't remember the error exactly (it was months ago) and I don't know how it got fixed. Maybe I did pushe something to fix it or someone else did, but I just don't remember anymore :-)

[avnik]

On Wed, Jan 20, 2016 at 01:14:07PM -0800, Vladimír Čunát wrote:

It didn't build, but I don't remember exactly (it was months
ago) and I don't know how they got fixed. Maybe I did pushed
something to fix it or someone else did, but I just don't
remember anymore :-)

Or it just come from cache.
I will try to invalidate my local build (by adding space
somewhere) and look if it build or not.

[dezgeg]

Probably the subversion problem was caused by GCC 5 (which closure-size had but no longer seems to have) which had some changes in that area.

[avnik]

On Mon, Jan 25, 2016 at 03:51:01PM -0800, Tuomas Tynkkynen wrote:

Probably the subversion problem was caused by GCC 5 (which closure-size had but no longer seems to have) which had some changes in that area.

In my experiments it fixed by adding

CPPFLAGS="-P";

But not sure if it correct solution.

[avnik]

On Mon, Jan 25, 2016 at 03:51:01PM -0800, Tuomas Tynkkynen wrote:

Probably the subversion problem was caused by GCC 5 (which closure-size had but no longer seems to have) which had some changes in that area.

Some comment taken from my own unstaged commit:

Mysterious error "configure: error: failed to recognize APR_INT64_T_FMT on this platform" appeared,
usually it appears on OSX, not on linux. Openembedded's forum suggest
CPPFLAGS="-P" to fix it.

Short snippet from svn's autocrap illustrating problematic place:
140           if test "x$svn_cv_pycfmt_apr_int64_t" = "x"; then
141             AC_EGREP_CPP([MaTcHtHiS +\"lld\" +EnDeNd],
142                          [#include <apr.h>
143                           MaTcHtHiS APR_INT64_T_FMT EnDeNd],
144                          [svn_cv_pycfmt_apr_int64_t="L"])
145           fi

I think line numbering can smash this magic, so disable it with `-P`

[vcunat]

The gcc switch back was done and explained in 9d13434. It didn't occur to me as a possible reason why it builds again.

[avnik]

Updated branch.
(remove svn stuff, fix build deps for syslog-ng)

[avnik]

@vcunat This one can be merged ;)

Next broken things: qtbase (via postgres's libpq and iodbc), and packages which contain only manpages (stdmanpage) -- but haven't any ideas how to fix them.

Any other packages which I can to fix/test to make whole closure-size merged?

[vcunat]

qtbase linking to libpq should be fixed by 363f667. I don't see a connection between qt and iodbc in the nix expressions.

[vcunat]

The stdman problem was that developer documentation (such as man3) is now stripped from packages by default or moved into .docdev if defined. I passed nixpkgs/pkgs/data/documentation files in c979012.

[avnik]

On Thu, Jan 28, 2016 at 04:39:18AM -0800, Vladimír Čunát wrote:

qtbase linking to libpq should be fixed by 363f667.
I don't see a connection between qt and iodbc in the nix
expressions.

I just see iodbc related error upper in log before libpq's one.

[vcunat]

I expect it's an optional dependency that is never fulfilled on any of our branches.

[avnik]

On Thu, Jan 28, 2016 at 06:16:14AM -0800, Vladimír Čunát wrote:

I expect it's an optional dependency that is never fulfilled
on any of our branches.

Good.
@vcunat have we any issue(-s) with pending tasks to merge closure-size into master?

[vcunat]

I'm updating merge blockers (in my opinion) at the top of that PR #7701.