metaconfig.git
2 weeks agoZero out high bytes of 80-bit ldnan master
H.Merijn Brand - Tux [Wed, 23 Nov 2016 15:32:39 +0000 (16:32 +0100)] 
Zero out high bytes of 80-bit ldnan

1. Backport of dd6885360b15e763e0706dc28555eea328de432d

   fix garbage filtering with 80-bit long doubles

   The test had several problems that resulted in the excess
   bytes not getting zeroed out. This caused random contents in
   $Config{longdblinfbytes}, observed on Debian with GCC 6.2.0 (but not
   5.4.1).

   Bug-Debian: https://bugs.debian.org/844752

2. Backport of 6b2c7479d6d33249c2385d8a36d8ee9e61e1a519

   also zero out high bytes of 80-bit ldnan

   These are currently zero anyway, but things are probably not guaranteed
   to stay so.

3 weeks agoExtra note that there is NO need to install dist/meta from github
H.Merijn Brand - Tux [Mon, 14 Nov 2016 12:22:29 +0000 (13:22 +0100)] 
Extra note that there is NO need to install dist/meta from github

3 weeks agoAssume dist-git will merge our PR's
H.Merijn Brand - Tux [Sun, 13 Nov 2016 14:55:09 +0000 (15:55 +0100)] 
Assume dist-git will merge our PR's

3 weeks agoCleanups on files being worked on to be PR'd by dist-git
H.Merijn Brand - Tux [Sun, 13 Nov 2016 14:52:05 +0000 (15:52 +0100)] 
Cleanups on files being worked on to be PR'd by dist-git

3 weeks agoExclude diffs the will never be resolved
H.Merijn Brand - Tux [Sun, 13 Nov 2016 14:42:52 +0000 (15:42 +0100)] 
Exclude diffs the will never be resolved

3 weeks agoIndicate clean state
H.Merijn Brand - Tux [Sun, 13 Nov 2016 13:06:00 +0000 (14:06 +0100)] 
Indicate clean state

In this case we'd just have to wait for upstream to accept the PR

3 weeks agopromote dist-git rm_try changes to dist/
H.Merijn Brand - Tux [Sun, 13 Nov 2016 12:50:07 +0000 (13:50 +0100)] 
promote dist-git rm_try changes to dist/

3 weeks agoNo changes left to keep
H.Merijn Brand - Tux [Sun, 13 Nov 2016 12:49:28 +0000 (13:49 +0100)] 
No changes left to keep

3 weeks agoAllow a pattern for diff matching
H.Merijn Brand - Tux [Sun, 13 Nov 2016 11:26:41 +0000 (12:26 +0100)] 
Allow a pattern for diff matching

ls-diff.pl           => 161 files
ls-diff.pl -p rm_try => 40 files

4 weeks agoMissing from previous backport
H.Merijn Brand - Tux [Fri, 11 Nov 2016 15:52:40 +0000 (16:52 +0100)] 
Missing from previous backport

4 weeks agoExclude DOT from @INC - as agreed on for now
H.Merijn Brand - Tux [Fri, 11 Nov 2016 15:52:15 +0000 (16:52 +0100)] 
Exclude DOT from @INC - as agreed on for now

4 weeks agoConfigure: add defs summarizing doublekind/longdblkind
H.Merijn Brand - Tux [Fri, 11 Nov 2016 13:40:53 +0000 (14:40 +0100)] 
Configure: add defs summarizing doublekind/longdblkind

Backport of 4cb05021f1602b1c29295f791d76ba4b38426f2a

For windows/netware It seems that many of the recent fp definitions
have not been yet copied over there [1] [2], so went mostly by dead
reckoning [3].

[1] Note that many of them are not absolutely necessary for building.
[2] The proper updating involves doing stuff in win32, which I do not have.
[3] As far as I can tell, Windows CE does not really not have long double.

6 weeks agoAdd option to exclude '.' from @INC
H.Merijn Brand - Tux [Sat, 22 Oct 2016 12:05:33 +0000 (14:05 +0200)] 
Add option to exclude '.' from @INC

6 weeks agoAdd scan for IBM and Cray mainframe fp formats.
H.Merijn Brand - Tux [Sat, 22 Oct 2016 12:00:06 +0000 (14:00 +0200)] 
Add scan for IBM and Cray mainframe fp formats.

Backport of 6e6b38aecfbcb229c8568366e3566d7fe7c8bbc5
For completeness: it's quite unlikely Perl would build in those environments.

Though with Cray it's less impossible: Perl used to build in C90
UNICOS, in 5.8-ish timeframe.

With IBM, highly unlikely, because there probably never was a UNIXy
enough environment where the IBM Floating Point Architecture was used.

6 weeks agosignbit scan assuming too much
H.Merijn Brand - Tux [Sat, 22 Oct 2016 10:07:00 +0000 (12:07 +0200)] 
signbit scan assuming too much

Backport of b76dad153351f50f46ba283bb0da075b93e89838
It was assuming a negative zero, which is an IEEE-754 only concept.

There is no need to assume the negative zero for the correct
functioning of the signbit, however.

3 months agoAdd probe for gai_strerror
H.Merijn Brand - Tux [Thu, 11 Aug 2016 16:26:42 +0000 (18:26 +0200)] 
Add probe for gai_strerror

4 months agoConfigure: clarify "version-specific"
H.Merijn Brand - Tux [Thu, 28 Jul 2016 15:37:13 +0000 (17:37 +0200)] 
Configure: clarify "version-specific"

Backport of commit 6baaafe34b15a5bbc05abd45b1b00f684654bb25
Author: Dan Collins <dcollinsn@gmail.com>
Date:   Mon Jul 4 15:21:05 2016 -0400

    Configure: clarify "version-specific"

    This has long been a confusing configure question. It sounds like
    it saves time by installing /less/, but really it just installs
    binaries with a version number suffix.

4 months agoVAX Float format H
H.Merijn Brand - Tux [Thu, 28 Jul 2016 14:59:08 +0000 (16:59 +0200)] 
VAX Float format H

Backport of commits

Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Wed Jul 27 07:04:02 2016 -0400

    Configure: note that the infnan tests may crash.

    (And in VAX, that is exactly what happens with the infinities.)

commit 86ea01eb2de6e15e79ff54031d7fabfb5f628d4e
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Tue Jul 26 22:10:42 2016 -0400

    Configure: VAX format H

commit a0cc6aaf62d581814737e0f0869cd9ee9d8f88ef
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Mon Jul 25 21:50:32 2016 -0400

    VAX: catch vax floats beyond VMS or Ultrix

commit e90f85309441678b2b41bb1acd37873f9443bb67
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Mon Jul 25 18:57:41 2016 -0400

    VAX: document also formats S/T/X, and hidden bits

    Though S, T, and X are not really pure VAX formats per se.

5 months agoVAX: Configure changes for VAX floats
H.Merijn Brand - Tux [Tue, 5 Jul 2016 20:46:05 +0000 (22:46 +0200)] 
VAX: Configure changes for VAX floats

Backport of c183cd86045c09fcbba056a606ae50f11c9c5b5a

Detect the VAX floating point formats D and G.

And the F float, but that is float (duh), never likely to be
the double, but do it for consistency (we detect IEEE single
precision floats, too).

The T float and X float are the IEEE 64-bit and 128-bit,
but those were available only on the Alpha.

Tested on vax-netbsd.

5 months ago[perl #128538] [PATCH] Fix copy/paste error in Configure
H.Merijn Brand - Tux [Tue, 5 Jul 2016 16:47:13 +0000 (18:47 +0200)] 
[perl #128538] [PATCH] Fix copy/paste error in Configure

6 months agoNew probe for strerror_l() querylocale
Aaron Crane [Tue, 31 May 2016 18:42:50 +0000 (19:42 +0100)] 
New probe for strerror_l()

Requested by Karl Williamson for Perl.

This function is specified by POSIX.1-2008 (and therefore by Single Unix):

http://pubs.opengroup.org/onlinepubs/9699919799/functions/strerror_l.html

6 months agoNew probe for querylocale()
Aaron Crane [Fri, 22 Apr 2016 17:01:19 +0000 (18:01 +0100)] 
New probe for querylocale()

Requested by Karl Williamson for Perl.

This function is known to exist on FreeBSD and Mac OS, for example:

https://www.freebsd.org/cgi/man.cgi?query=querylocale
https://developer.apple.com/library/ios/documentation/System/Conceptual/ManPages_iPhoneOS/man3/querylocale.3.html

6 months agodist/meta merge after acceptance upstream
H.Merijn Brand - Tux [Sun, 29 May 2016 16:52:25 +0000 (18:52 +0200)] 
dist/meta merge after acceptance upstream

Three down, 161 to go

6 months agoFinish.U isn't modified anymore
H.Merijn Brand - Tux [Sun, 29 May 2016 13:30:33 +0000 (15:30 +0200)] 
Finish.U isn't modified anymore

All changes accepted by upstream. One down, 164 to go

6 months agoSpelling in dist installation
H.Merijn Brand - Tux [Sat, 14 May 2016 16:57:35 +0000 (18:57 +0200)] 
Spelling in dist installation

this change has been put as a PR to dist

6 months agoRemove trailing whitespace on meta-lines in unit files
H.Merijn Brand - Tux [Sat, 14 May 2016 16:48:26 +0000 (18:48 +0200)] 
Remove trailing whitespace on meta-lines in unit files

This change has also been proposed as PR to dist upstream

6 months agoConfused? You won't be after the next version of pigs in space
H.Merijn Brand - Tux [Sat, 14 May 2016 16:32:23 +0000 (18:32 +0200)] 
Confused? You won't be after the next version of pigs in space

I had to think about what the letters meant on every command, so they
were not logical enough. Now with comments in --help

6 months agoMerged changes to Finish.U
H.Merijn Brand - Tux [Sat, 14 May 2016 16:30:03 +0000 (18:30 +0200)] 
Merged changes to Finish.U

6 months agoTweak Csym.U probe to handle GCC 6 -flto
Aaron Crane [Sat, 14 May 2016 15:26:22 +0000 (16:26 +0100)] 
Tweak Csym.U probe to handle GCC 6 -flto

In GCC 6, link-time optimization is smarter at removing references to
functions that never actually get used, so all symbols were being detected
as available. This change constitutes our latest move in this arms race.

Backport of this perl5 commit:

commit adec5bbf0b66ee5ebc2ba80eda2389bb68e23d86
Author:     Lukas Mai <l.mai@web.de>
AuthorDate: Wed May 11 22:15:34 2016 +0200
Commit:     Aaron Crane <arc@cpan.org>
CommitDate: Sat May 14 16:21:36 2016 +0100

    fix symbol detection with gcc 6 link-time optimization (RT #128131)

6 months agoDon't croak when usecrosscompile but no targethost
Aaron Crane [Sat, 14 May 2016 13:15:56 +0000 (14:15 +0100)] 
Don't croak when usecrosscompile but no targethost

Backport of this perl5 commit, for Perl RT#127234:

commit 23695c073f41fdac9d20d873827ba8b2a3dda710
Author:     Niko Tyni <ntyni@debian.org>
AuthorDate: Sun Jan 10 22:35:53 2016 +0200
Commit:     Tony Cook <tony@develop-help.com>
CommitDate: Tue May 10 10:46:13 2016 +1000

    Fix the Configure escape with usecrosscompile but no targethost

    Commit 97076f2 added a graceful exit to Configure when targethost is
    not defined but usecrosscompile is. However, this is not reached because
    there is a similar check a bit earlier that makes Configure bail out.

    Make the earlier check warn instead of croaking.

    The use case for this combination is supplying an external config.sh
    suitable for the target platform, avoiding the need for configuration
    probes.

7 months agoBring more dist in line with dist-git
H.Merijn Brand - Tux [Sat, 23 Apr 2016 17:08:19 +0000 (19:08 +0200)] 
Bring more dist in line with dist-git

All of these still have been overruled units, so it'll change nothing
in functionality but should ease finding diffs that do matter

7 months ago[UTIL] Perl system does not default to $_
Aaron Crane [Sat, 23 Apr 2016 10:30:54 +0000 (11:30 +0100)] 
[UTIL] Perl system does not default to $_

This doesn't actually affect anything, because no units have a ?MAKE rule
with any actions other than the "pick" pseudo-command. But it seems worth
fixing nonetheless.

7 months ago[UTIL] mconfig: don't rely on directory order
Aaron Crane [Sat, 23 Apr 2016 10:30:06 +0000 (11:30 +0100)] 
[UTIL] mconfig: don't rely on directory order

7 months agoUpdate README for FS type and mconfig
H.Merijn Brand - Tux [Sat, 23 Apr 2016 10:50:46 +0000 (12:50 +0200)] 
Update README for FS type and mconfig

7 months agoremoved but saved old dist folder
H.Merijn Brand - Tux [Sat, 23 Apr 2016 10:32:05 +0000 (12:32 +0200)] 
removed but saved old dist folder

7 months agoOne down, too many to go arc/tmp-tux
H.Merijn Brand - Tux [Fri, 22 Apr 2016 17:05:49 +0000 (19:05 +0200)] 
One down, too many to go

7 months agoMass update of *headers* in dist/U from git
H.Merijn Brand - Tux [Fri, 22 Apr 2016 16:51:31 +0000 (18:51 +0200)] 
Mass update of *headers* in dist/U from git

This is to simplify identifying upcoming changes

7 months agounused leftover
H.Merijn Brand - Tux [Fri, 22 Apr 2016 16:13:39 +0000 (18:13 +0200)] 
unused leftover

7 months agogit => dist
H.Merijn Brand - Tux [Fri, 22 Apr 2016 16:09:38 +0000 (18:09 +0200)] 
git => dist

These are the simple copies with no changes but the RCS headers

7 months agoUpdate dist/U from dist-git/mcon/U
H.Merijn Brand - Tux [Fri, 22 Apr 2016 15:13:31 +0000 (17:13 +0200)] 
Update dist/U from dist-git/mcon/U

First step.

perl ls-diff.pl will show

  #     Git             Dist              Perl    Diff Unit
=== ========= ====== ========= ====== ========= ====== ======================
  1  3974/135   1113  4014/135   4732  3804/138   4687 Checkcc.U
  2  2327/ 63    696  2372/ 63   2221  2182/ 61   2176 Compile.U
  3  4076/114   2496  4161/114   2463  4153/111   3289 Config_h.U
:
:

size of the unit/number of lines in the unit size of the diff
compared to the next column. Last diff is the size of the diff
between dist-git and modified.

Next steps will include updating dist/U from dist-git/mcon/U
followed by the same updates from dist/U into U (our modified
versions). That last step should either remove our modified
version or result in feedback to Rafael in order to explain
what we changed and why, hoping it'll find its way upstream so
we can remove yet another modified unit not to maintain :)

Expect a flurry of commits to dist/U where only comments and
typoes are fixed.

7 months agoRemove a module that has no changes to the original from dist
H.Merijn Brand - Tux [Fri, 22 Apr 2016 14:41:16 +0000 (16:41 +0200)] 
Remove a module that has no changes to the original from dist

The changes are merely in the RCS headers and these do not conflict
Replace the dist unit with the one from dist-git

7 months ago[UTIL] New metagrep -F option
Aaron Crane [Fri, 22 Apr 2016 12:44:46 +0000 (13:44 +0100)] 
[UTIL] New metagrep -F option

Does fixed-string matching, like grep -F.

7 months agoProbe for and expose the 'fd' field for SA_SIGINFO
Aaron Crane [Fri, 22 Apr 2016 09:33:33 +0000 (10:33 +0100)] 
Probe for and expose the 'fd' field for SA_SIGINFO

Backport of this commit, which is currently on the ilmari/siginfo-fd branch
in perl5.git:

commit 95288b852f4f3a4e7b82ecd37ae330943af64340
Author: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Date:   Wed Feb 3 16:35:15 2016 +0000

    Probe for and expose the 'fd' field for SA_SIGINFO

    This Linux-specific field indicates which FD a SIGIO/SIGPOLL signal
    pertains to.

7 months agoadd d_duplocale and i_locale Configure probes to d_newlocale.U
Aaron Crane [Fri, 22 Apr 2016 09:20:04 +0000 (10:20 +0100)] 
add d_duplocale and i_locale Configure probes to d_newlocale.U

Backport of this commit, which is currently on the smoke-me/khw-new_locale
branch in perl5.git:

commit 954392da8720867ead32088bd3483655b6623428
Author: Tony Cook <tony@develop-help.com>
Date:   Mon Mar 21 12:12:58 2016 +1100

    add d_duplocale and i_locale Configure probes

7 months agocosmetic doc change
H.Merijn Brand - Tux [Fri, 22 Apr 2016 07:47:30 +0000 (09:47 +0200)] 
cosmetic doc change

7 months agoFix wording for glossary of lrint() and related functions
Aaron Crane [Thu, 21 Apr 2016 16:20:04 +0000 (17:20 +0100)] 
Fix wording for glossary of lrint() and related functions

This corrects infelicities in the following:

commit bc580dd327b37b38d4f66d5c335b3e9fc12eaac8
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Mon Feb 2 17:26:16 2015 -0500

    Configure: lrintl, lroundl, llrintl, llroundl

7 months agoDelete Configure probe for d_modfl_pow32_bug in d_modfl.U
Aaron Crane [Thu, 21 Apr 2016 16:09:08 +0000 (17:09 +0100)] 
Delete Configure probe for d_modfl_pow32_bug in d_modfl.U

Perl commit 94a2587fdd5c930411410a33daa1209d3ef941ea removed the glossary
entries and the like, but left the Configure probe in place.

7 months agoAdd missing ?H: lines for d_longdbl.U
Aaron Crane [Thu, 21 Apr 2016 15:23:40 +0000 (16:23 +0100)] 
Add missing ?H: lines for d_longdbl.U

This should have been part of 9c52d4117799c8de73f4871a148d3bb88fe5214d

7 months agoFix 683e3b31e8b72cce
H.Merijn Brand - Tux [Thu, 21 Apr 2016 14:43:46 +0000 (16:43 +0200)] 
Fix 683e3b31e8b72cce

Was committed in perl5 as a1da3d4009387b6065a5116ba3a6cccd85e70168
Author: H.Merijn Brand <h.m.brand@xs4all.nl>
Date:   Sat Oct 10 15:27:09 2015 +0200

    Found when backporting. JHI's reply:

    Oops.

    It should be define:define:define.

    I was trying to stop the test that uses alarm() not to compiled in a
    system where alarm() does not exist.

7 months agoThese two are not required in perl, but got included anyway
H.Merijn Brand - Tux [Thu, 21 Apr 2016 14:33:05 +0000 (16:33 +0200)] 
These two are not required in perl, but got included anyway

We can alway get them from diat-git if we would need them later

7 months agoDon't add -shared when the compiler is clang++
H.Merijn Brand - Tux [Thu, 21 Apr 2016 14:13:14 +0000 (16:13 +0200)] 
Don't add -shared when the compiler is clang++

Backport of f8fbb01c4686d7e4393ba90b2e1326e9536d9107
Author: Chris 'BinGOs' Williams <chris@bingosnet.co.uk>
Date:   Thu Aug 27 15:30:10 2015 +0100

    Don't add -shared when the compiler is clang++

7 months ago[UTIL] New simplifydiff tool
Aaron Crane [Thu, 21 Apr 2016 14:11:05 +0000 (15:11 +0100)] 
[UTIL] New simplifydiff tool

Pipe a diff into it, and it removes any diff hunk in which every insertion
was deleted by some other part of the diff, and every deletion was inserted
by some other part.

7 months agoVarious updates to float-format probes
Aaron Crane [Thu, 21 Apr 2016 14:06:25 +0000 (15:06 +0100)] 
Various updates to float-format probes

Backported from the following:

commit b5897239a0c3c2f5ec699690086b30e7db4305d4
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Tue Dec 8 08:40:38 2015 -0500

    Configure: cannot trust the bytes after the 80-bit fp

    They can be zero, they can be garbage.

commit 3118d7d684b56cbeb702af874f4326683c45f045
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Tue Dec 8 08:48:40 2015 -0500

    Configure: mixed-endian double-doubles

    The ppc64el is the first seen little-endian double-double (and also
    the first little-endian ppc), but it turns out its little-endianness
    is mixed: the doubles are still in big-endian order.  Configure was
    expecting wrongly a fully byte-reversed double-double.

    Therefore extend the long double format detection to cover all the
    (double-double) permutations, though the formats of five and eight
    are rather unlikely (based on current platforms using double-double).

commit 74c6ce8743ba719b8416ce00933e19104484b949
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Thu Dec 10 18:59:53 2015 -0500

    Configure: infnan for mixed-endian double-double

commit 157a0ac7a2e2f56bd291b87c343712def4a1acfb
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Wed Dec 16 21:57:31 2015 -0500

    Configure: notes on the m68881 extended precision format

7 months agoamigaos4: Configure: syslog extension not feasible
H.Merijn Brand - Tux [Thu, 21 Apr 2016 14:04:26 +0000 (16:04 +0200)] 
amigaos4: Configure: syslog extension not feasible

Backport of 57bce8812ba4e6da4ddb13a5a2340d3329ac6cd5
Author: Andy Broad <andy@broad.ology.org.uk>
Date:   Mon Sep 14 10:37:45 2015 -0400

    amigaos4: Configure: syslog extension not feasible

7 months agodead symlink
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:58:16 +0000 (15:58 +0200)] 
dead symlink

7 months agoGet rid of the hasfield_t global variable
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:58:02 +0000 (15:58 +0200)] 
Get rid of the hasfield_t global variable

7 months agoTry to avoid doubling the cppflags if rerun without removing config.sh
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:32:27 +0000 (15:32 +0200)] 
Try to avoid doubling the cppflags if rerun without removing config.sh

Backport of 76e6b6c7dc4c93d421dfe6c24866c3cd4560b57a
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Fri Oct 16 08:10:40 2015 -0400

    Try to avoid doubling the cppflags if rerun without removing config.sh

    This bug has been here since 2001, introduced by yours truly
    in 58e77565.  Hasn't been too harmful, obviously.  It doubles
    the cppflags only once, thankfully, unlimited doubling would
    probably have been noticed earlier.

    The avoidance maneuver is far from fool-proof.  To be more
    fooler-proofer, some sort of order-preserving deduping would
    be needed.

7 months agoAnother small cleanup
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:31:23 +0000 (15:31 +0200)] 
Another small cleanup

7 months agoRemove old cruft. It still is in git
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:24:57 +0000 (15:24 +0200)] 
Remove old cruft. It still is in git

7 months agocompile error after re-running Configure since AmigaOS merge
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:24:30 +0000 (15:24 +0200)] 
compile error after re-running Configure since AmigaOS merge

Backport of b44694539e922b1078113e05f80651c9eedd5d37
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Thu Oct 15 08:39:42 2015 -0400

    rt.perl.org 126152 compile error after re-running Configure since AmigaOS merge

    Configure /proc issues, honor d_procselfexe and procselfexe hints.

7 months agoStuff missed in previous commits
H.Merijn Brand - Tux [Thu, 21 Apr 2016 13:17:23 +0000 (15:17 +0200)] 
Stuff missed in previous commits

7 months agoMake -O behavior the default
H.Merijn Brand - Tux [Thu, 21 Apr 2016 12:04:43 +0000 (14:04 +0200)] 
Make -O behavior the default

Backport of 41d73075f0801c26794dadb1ff690f305d7e53a7
Author: Ivan Pozdeev <vano@mail.mipt.ru>
Date:   Wed Aug 12 20:33:12 2015 +0300

and the later update

commit 7a4fcb3f64df51eacf89f69a4bf76386aea1e8e3
Author: Aaron Crane <arc@cpan.org>
Date:   Wed Dec 2 00:06:42 2015 +0000

    Configure: unbreak -S option now that -O is the default

    As far as I can tell, using the -S and -O options together has always
    yielded an error of this form:

        Configure: 2042: .: Can't open ./optdef.sh

    That's because, even though optdef.sh is created in the UU directory, and
    most of Configure is run in that directory, part of the -S implementation is
    run in the root directory, and was therefore trying to read ./optdef.sh
    instead of ./UU/optdef.sh.

    As of 41d73075f0801c26794dadb1ff690f305d7e53a7, the -O mode is always
    enabled, so the -S option has been broken since then. This fixes that.

7 months agoprobe in Configure whether dtrace builds an object
Aaron Crane [Thu, 21 Apr 2016 11:55:30 +0000 (12:55 +0100)] 
probe in Configure whether dtrace builds an object

backport of:

commit c2538af7458bf317cdc7bc684f65831744010d80
Author: Tony Cook <tony@develop-help.com>
Date:   Wed Mar 9 11:54:13 2016 +1100

    [perl #122287] probe in Configure whether dtrace builds an object

    When building the object file, newer versions of dtrace (on Illumos
    based systems at least) require an input object file that uses
    at least one of the probes defined in the .d file.

    The test in Makefile.SH didn't provide that definition so the test
    would fail, and not build an object file, and fail to link later on,
    on systems that *do* need the object file.

    Moved the probe to Configure (where it probably belongs) and supplied
    an object file that uses a probe.

    Tested successfully on OmniOS (with the new dtrace), Solaris 11,
    and darwin.

commit 5fa8e144165a5c086facccf35630b9c4a781c4ad
Author: David Mitchell <davem@iabyn.com>
Date:   Fri Mar 18 23:29:44 2016 +0000

    Configure: silence 'dtrace -G' probe

    When built with -Dusedtrace, Configure checks to see whether 'dtrace -G'
    is supported, by running it. If it fails, it may spew error messages
    to stderr, so use >/dev/null 2>&1.

7 months ago[UTIL] Scan each file max once
H.Merijn Brand - Tux [Thu, 21 Apr 2016 11:44:41 +0000 (13:44 +0200)] 
[UTIL] Scan each file max once

7 months agoIt does not work in SysV (solaris) or old BSD greps.
H.Merijn Brand - Tux [Thu, 21 Apr 2016 11:35:36 +0000 (13:35 +0200)] 
It does not work in SysV (solaris) or old BSD greps.

Even though -q is POSIX, that does not guarantee succes

Backport of e244340e4067bb332773529bbff797bd14f103de
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Wed Dec 16 20:19:03 2015 -0500

    Configure: grep -q is not portable

    It does not work in SysV (solaris) or old BSD greps.

7 months agosiginfo_si part
H.Merijn Brand - Tux [Thu, 21 Apr 2016 11:30:09 +0000 (13:30 +0200)] 
siginfo_si part

Backport of commit 6e72ca7018e96d54b93ca8c41f8033a1f9c2e874
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Tue Mar 1 19:56:04 2016 -0500

    amigaos4: no sigaction, no si fields

7 months agosilence some try.c warnings
H.Merijn Brand - Tux [Thu, 21 Apr 2016 11:18:45 +0000 (13:18 +0200)] 
silence some try.c warnings

Backport from cc0bf92f2c5d6a67a850efa7d21a5a83e016d446
Author: David Mitchell <davem@iabyn.com>
Date:   Mon Mar 14 16:49:05 2016 +0000

    Configure: silence some try.c warnings

    On one of the try.c compilations, redirect stderr to /dev/null, since
    the code can legitimately warn without there being a problem.

    The try.c in question is probing for what symbols the compiler supports.
    The clang extensions __has_include and __has_include_next are designed
    only to be used in .h files, so they warn if used from try.c

7 months agoConfigure: gcc 5 exists.
H.Merijn Brand - Tux [Thu, 21 Apr 2016 11:12:14 +0000 (13:12 +0200)] 
Configure: gcc 5 exists.

Backport of 11c0dec80e86a2efeef002853ab7f0e12199f729
Author: Jarkko Hietaniemi <jhi@iki.fi>
Date:   Sun Nov 15 19:45:06 2015 -0500

    Configure: gcc 5 exists.

    (cherry picked from commit 0632167b5a653661f48c9e38df6aa76b544d4079)

7 months agoProbe for and expose more fields for SA_SIGINFO
H.Merijn Brand - Tux [Thu, 21 Apr 2016 10:58:11 +0000 (12:58 +0200)] 
Probe for and expose more fields for SA_SIGINFO

Backport of cdfe2e65ec1960d301d083bee421388e064cd234
Author: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Date:   Tue Jan 12 14:47:07 2016 +0000

    Probe for and expose more fields for SA_SIGINFO

    These are all specified by POSIX/SUSv3, but not all platforms have them,
    as mentioned in POSIX.pm.

    We can only test the pid, uid and code fields, since they are the only
    ones that are defined for a user-sent signal.

7 months agoPorting/Glossary: fix a set of typos
Aaron Crane [Thu, 21 Apr 2016 09:57:21 +0000 (10:57 +0100)] 
Porting/Glossary: fix a set of typos

A few descriptions of floating-point formats included the word "big" before
the actual endianness.

backport of aaca660c731cd2a7c43471241fd380e082a05f57

7 months agoensure alignbytes is correct for -Dusequadmath
Aaron Crane [Thu, 21 Apr 2016 09:12:46 +0000 (10:12 +0100)] 
ensure alignbytes is correct for -Dusequadmath

backport of ab462c7d2da3570ffe26e717642be96cccef8b23

7 months agoRemove nm from libswanted
H.Merijn Brand - Tux [Wed, 20 Apr 2016 19:26:10 +0000 (21:26 +0200)] 
Remove nm from libswanted

Backport of 9a427118a27dce49b50e2c7d619b9f5aa048101d

8 months agodist/meta moved to github \o/ \o/
H.Merijn Brand - Tux [Fri, 1 Apr 2016 10:07:04 +0000 (12:07 +0200)] 
dist/meta moved to github \o/ \o/

8 months agoAdd probe for memmem
H.Merijn Brand - Tux [Sat, 26 Mar 2016 08:31:23 +0000 (09:31 +0100)] 
Add probe for memmem

8 months agoAdd Configure probes for newlocale, freelocale, and uselocale
H.Merijn Brand - Tux [Sat, 19 Mar 2016 14:37:43 +0000 (15:37 +0100)] 
Add Configure probes for newlocale, freelocale, and uselocale

11 months agoAdd unit to detect fdclose(), introduced in FreeBSD.
Andy Dougherty [Sun, 3 Jan 2016 22:49:34 +0000 (17:49 -0500)] 
Add unit to detect fdclose(), introduced in FreeBSD.

14 months agoBackporting work on AmigaOS
H.Merijn Brand - Tux [Sat, 10 Oct 2015 13:24:49 +0000 (15:24 +0200)] 
Backporting work on AmigaOS

With an additional fix for bad commit 683e3b31e8b72ccea0ba577bdee2bc58caa738ad

Serious problem with the change to Configure itself is that a case/esac
was added that span TWO metaunits, so both had to get it. I don't know
if adding the LINT:extern in Signal.U is safe enough

16 months agoConsistently use -o a.out (spotted by Ivan Pozdeev)
H.Merijn Brand - Tux [Thu, 30 Jul 2015 08:46:03 +0000 (10:46 +0200)] 
Consistently use -o a.out (spotted by Ivan Pozdeev)

16 months agoThere is no default dflt for usedtrace (spotted by Ivan Pozdeev)
H.Merijn Brand - Tux [Thu, 30 Jul 2015 08:30:30 +0000 (10:30 +0200)] 
There is no default dflt for usedtrace (spotted by Ivan Pozdeev)

17 months agodescription swap
H.Merijn Brand - Tux [Mon, 6 Jul 2015 13:55:17 +0000 (15:55 +0200)] 
description swap

17 months agoTwo nits not show by mlint
H.Merijn Brand - Tux [Mon, 6 Jul 2015 07:38:40 +0000 (09:38 +0200)] 
Two nits not show by mlint

17 months agoinfnan: move the mantbits definitions from perl.h to Configure
H.Merijn Brand - Tux [Mon, 6 Jul 2015 07:28:27 +0000 (09:28 +0200)] 
infnan: move the mantbits definitions from perl.h to Configure

Backport of 5a14060af165c24c63cfc5d5f4d2aeb52cc41a03

17 months agoinfnan: Configure scan for fp mantissa bytes
H.Merijn Brand - Tux [Mon, 6 Jul 2015 07:22:23 +0000 (09:22 +0200)] 
infnan: Configure scan for fp mantissa bytes

backport of ed3917fd69b234bb5614cb9aed93d62238e3dcb8

17 months agoinfnan: Configure scan for infnan bytes
H.Merijn Brand - Tux [Mon, 6 Jul 2015 07:08:55 +0000 (09:08 +0200)] 
infnan: Configure scan for infnan bytes

backport 44521f3a1782026b7d25cc55af459c3e28cc9bdd with fixes

17 months agoCreate try.c afresh for pointersize test
H.Merijn Brand - Tux [Mon, 6 Jul 2015 06:30:56 +0000 (08:30 +0200)] 
Create try.c afresh for pointersize test

19 months agoInhibit the use of \ in -D arguments
H.Merijn Brand - Tux [Sun, 19 Apr 2015 08:56:12 +0000 (10:56 +0200)] 
Inhibit the use of \ in -D arguments

21 months agoConfigure: add note about nvsize sometimes lying
H.Merijn Brand - Tux [Fri, 6 Mar 2015 12:40:14 +0000 (13:40 +0100)] 
Configure: add note about nvsize sometimes lying

Lying in case you expect that many bytes being used for NV.

Backport of 0cd93aca4e274765289e9e01d34c4a5dcf06df1c
Author: Jarkko Hietaniemi <jhi@iki.fi>  2015-03-02 13:00:33

21 months agoMissing Configure messages.
H.Merijn Brand - Tux [Fri, 6 Mar 2015 12:33:17 +0000 (13:33 +0100)] 
Missing Configure messages.

(The scans have been happening, the echo case branches were missing.)

Backport of e11fa374c86b187ae1e8382680d49e2e44abf1bb
Author: Jarkko Hietaniemi <jhi@iki.fi>  2015-02-25 13:51:53

21 months agouname -r works on MINIX as well
H.Merijn Brand - Tux [Mon, 16 Feb 2015 09:07:12 +0000 (10:07 +0100)] 
uname -r works on MINIX as well

Backport of 0cf583b0808e2fb2ac90751d506f2b8ff75aed86

21 months agoAdd lrintl, llrintl, lroundl, and llroundl probes
H.Merijn Brand - Tux [Mon, 16 Feb 2015 08:59:43 +0000 (09:59 +0100)] 
Add lrintl, llrintl, lroundl, and llroundl probes

Backport of bc580dd327b37b38d4f66d5c335b3e9fc12eaac8 with some small
document changes

23 months agoPut pthread in front of libswanted and add cl
H.Merijn Brand - Tux [Mon, 5 Jan 2015 11:31:45 +0000 (12:31 +0100)] 
Put pthread in front of libswanted and add cl

If pthread is found on HP-UX, it is required to be in front and
libcl is required too

23 months agosilence ASan warnings
H.Merijn Brand - Tux [Wed, 31 Dec 2014 12:07:08 +0000 (13:07 +0100)] 
silence ASan warnings

Backport of
commit a5ba819fc61657341addc515e13a09c715b9597d
Author: David Mitchell <davem@iabyn.com>
Date:   Mon Dec 22 21:36:14 2014 +0000

    Configure: silence ASan warnings

    When run under -fsanitize=undefined, some of the try.c's that are compiled
    and executed give runtime warnings. Since the intent of these particular
    executables is to probe beyond certain limits in order to determine those
    limits, these warnings can be safely ignored.  So file them in /dev/null.

23 months agoPerl 5 on Haiku - libperl.so not found installing out of standard location
H.Merijn Brand - Tux [Wed, 24 Dec 2014 20:07:47 +0000 (21:07 +0100)] 
Perl 5 on Haiku - libperl.so not found installing out of standard location

Patch by Leon, checked by Dan Collins

23 months agoadd pthread to libswanted
H.Merijn Brand - Tux [Thu, 11 Dec 2014 17:25:55 +0000 (18:25 +0100)] 
add pthread to libswanted

2 years agosilence debug output (from v5.21.6-239-g31b42f2)
H.Merijn Brand - Tux [Wed, 3 Dec 2014 13:56:48 +0000 (14:56 +0100)] 
silence debug output (from v5.21.6-239-g31b42f2)

2 years agoGlossary updates
H.Merijn Brand - Tux [Sun, 23 Nov 2014 17:27:40 +0000 (18:27 +0100)] 
Glossary updates

d_cbrt also changed upstream

2 years agoGet in the last bunch of Configure changes
H.Merijn Brand - Tux [Thu, 20 Nov 2014 15:21:57 +0000 (16:21 +0100)] 
Get in the last bunch of Configure changes

Also bring config_h.SH in sync

2 years agoBackport all of Jarkko's new (math) function probes
H.Merijn Brand - Tux [Tue, 18 Nov 2014 21:14:15 +0000 (22:14 +0100)] 
Backport all of Jarkko's new (math) function probes

next are the other changes and checks on Glossary and config_h.SH

2 years agoHeader change
H.Merijn Brand - Tux [Tue, 18 Nov 2014 14:16:06 +0000 (15:16 +0100)] 
Header change

Asked for more specific description by Debian
Agreed by Rafael (note the text above):
--8<---
Of course, that is just fine since a pointer to the original distribution
is kept, whilst mentioning the specific derived version Perl uses!

Thanks for double-checking though!

Raphael
-->8---

2 years agoAlso add all utilities for building from units to repo
H.Merijn Brand - Tux [Wed, 1 Oct 2014 06:08:08 +0000 (08:08 +0200)] 
Also add all utilities for building from units to repo

PLEASE do NOT modify these in the repo, as that will break MY builds.
I only moved them here to have all readily available

Note that some still need manual change if used locally
• mconfig
• metaconfig
• metagrep
• metalint
• metaxref
• mlint
• packinit
• patcil
• patpost
• patsend