Re: [PATCH] Makefile fix

[Doug Evans <xdje42 at gmail dot com>]

On Sun, Feb 9, 2014 at 4:58 PM, Doug Evans <xdje42@gmail.com> wrote:
> On Sat, Feb 8, 2014 at 3:41 AM, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>>> Date: Fri, 7 Feb 2014 21:00:35 -0800
>>> From: Doug Evans <dje@google.com>
>>>
>>> On Tue, Feb 4, 2014 at 2:33 AM, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>>> > The diff below fixes an issue that has been bothering me for quite a
>>> > while.  When using OpenBSD make, almost everything in gdb/ gets
>>> > rebuilt.  The problem is the use of the all-lib phony target, which is
>>> > always considered to be out-of-date, and used as a prerequisite of
>>> > libgnu.a and the gnulib string.h, which are therefore also considered
>>> > to be out-of-date.  I have no idea why this problem doesn't show up
>>> > with GNU make.  But the GNU make manual clearly states that phony
>>> > targets should not be prerequisites of real target files:
>>> >
>>> >   <http://www.gnu.org/software/make/manual/make.html#Phony-Targets>
>>> >
>>> > It seems to me that the phony target can be easily avoided; see the
>>> > diff below.  But perhaps I'm missing something?
>>> >
>>> >
>>> >
>>> > diff --git a/gdb/Makefile.in b/gdb/Makefile.in
>>> > index ed84e35..937478b 100644
>>> > --- a/gdb/Makefile.in
>>> > +++ b/gdb/Makefile.in
>>> > @@ -1263,10 +1263,8 @@ gdb$(EXEEXT): gdb.o $(LIBGDB_OBS) $(ADD_DEPS) $(CDEPS) $(TDEPLIBS)
>>> >                 $(TDEPLIBS) $(TUI_LIBRARY) $(CLIBS) $(LOADLIBES)
>>> >
>>> >  # Convenience rule to handle recursion.
>>> > -$(LIBGNU) $(GNULIB_H): all-lib
>>> > -all-lib: $(GNULIB_BUILDDIR)/Makefile
>>> > +$(LIBGNU) $(GNULIB_H): $(GNULIB_BUILDDIR)/Makefile
>>> >         @$(MAKE) $(FLAGS_TO_PASS) DO=all DODIRS=$(GNULIB_BUILDDIR) subdir_do
>>> > -.PHONY: all-lib
>>> >
>>> >  # Convenience rule to handle recursion.
>>> >  .PHONY: all-data-directory
>>>
>>> I found the patch where this went in but I couldn't find anything interesting.
>>> "all-lib" is there, I'm guessing, just as a convenience for typing,
>>> but I'm not sure how often it's used so if it's getting in the way I'm
>>> ok with this change.
>>> I can't think of any problems with the patch.
>>> [And we could still keep all-lib as a convenience rule, but I'm ok
>>> with deleting it for now.]
>>
>> Thanks.  Here's what I committed:
>>
>> From 0a6bd22df74bb5b6ffa30c3b10243a34d9055c32 Mon Sep 17 00:00:00 2001
>> From: Mark Kettenis <kettenis@gnu.org>
>> Date: Sat, 8 Feb 2014 12:36:17 +0100
>> Subject: [PATCH] A phony target should not be a prerequisite of a real target
>>  file
>>
>> gdb/ChangeLog:
>>
>>         * Makefile.in (all-lib): Remove.
>>         ($(LIBGNU) $(GNULIB_H)): Replace with gits of remove all-lib target.
>> ---
>>  gdb/ChangeLog   | 5 +++++
>>  gdb/Makefile.in | 4 +---
>>  2 files changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/gdb/ChangeLog b/gdb/ChangeLog
>> index c2c9236..bb64a23 100644
>> --- a/gdb/ChangeLog
>> +++ b/gdb/ChangeLog
>> @@ -1,3 +1,8 @@
>> +2014-02-08  Mark Kettenis  <kettenis@gnu.org>
>> +
>> +       * Makefile.in (all-lib): Remove.
>> +       ($(LIBGNU) $(GNULIB_H)): Replace with gits of remove all-lib target.
>> +
>>  2014-02-07  Doug Evans  <dje@google.com>
>>
>>         * extension-priv.h (extension_language_script_ops): Add comment.
>> diff --git a/gdb/Makefile.in b/gdb/Makefile.in
>> index e714550..0976615 100644
>> --- a/gdb/Makefile.in
>> +++ b/gdb/Makefile.in
>> @@ -1265,10 +1265,8 @@ gdb$(EXEEXT): gdb.o $(LIBGDB_OBS) $(ADD_DEPS) $(CDEPS) $(TDEPLIBS)
>>                 $(TDEPLIBS) $(TUI_LIBRARY) $(CLIBS) $(LOADLIBES)
>>
>>  # Convenience rule to handle recursion.
>> -$(LIBGNU) $(GNULIB_H): all-lib
>> -all-lib: $(GNULIB_BUILDDIR)/Makefile
>> +$(LIBGNU) $(GNULIB_H): $(GNULIB_BUILDDIR)/Makefile
>>         @$(MAKE) $(FLAGS_TO_PASS) DO=all DODIRS=$(GNULIB_BUILDDIR) subdir_do
>> -.PHONY: all-lib
>>
>>  # Convenience rule to handle recursion.
>>  .PHONY: all-data-directory
>> --
>> 1.8.5.3
>>
>
> I'm getting gnulib related build failures on linux, I'm guessing it's
> this change.
> Sorry 'bout that.
> I think it's that a parallel make is now entering build-gnulib twice.
> So we need to serialize it, without using a phony target.

I reverted the patch to get linux builds going.
Again., sorry 'bout that.

I'll work on a different fix on Monday, unless you want to take it.