I should have emphasized that there are no diffs for Automake's
Makefile.ins. The test suite is now at libobj*.test, no errors so far.
Tom Tromey <[EMAIL PROTECTED]> writes:
> Me, too, but:
>
> * Automake is really slow
> * We already have tons of other global constants
>
> Please change it back.
No problem!
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
>> How smart is Perl?
>> I used to have this as a global so that it wasn't reevaluated each
>> time through the loop. It never changes -- it is a constant
>> structure.
Akim> I privileged readability and modularity over efficiency.
Me,
Tom Tromey <[EMAIL PROTECTED]> writes:
> Akim> + # Directories installed during 'install-exec' phase.
> Akim> + my %exec_dir_p =
> Akim> + (
> Akim> [ ... ]
>
> How smart is Perl?
> I used to have this as a global so that it wasn't reevaluated each
> time through
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> * automake.in (&am_install_var): Recognize whether predefined
Akim> directories are `data' or `exec'.
Akim> * libs.am, ltlib.am, progs.am, python.am, scripts.am: Install
Akim> ?EXEC?.
You checked this in already, which is fi
Given that this time I think I understood, that the bug was really bad
for users, I went ahead and applied it.
Tom, feel free to strengthen instdata2!
Here is what I committed.
#! /bin/sh
# Test that installing under $exec_prefix is handled by install-exec.
# Testing with headers for instance.
I think I finally have it. I apologize for all the iterations I
needed :(
Index: ChangeLog
from Akim Demaille <[EMAIL PROTECTED]>
* automake.in (&am_install_var): Recognize whether predefined
directories are `data' or `exec'.
* libs.am, ltlib.am, progs.am, python.am, s
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> So it is my (today's :) understanding that I did wrong to have
Akim> some primaries independent from the dir name, right?
Right. The primaries are irrelevant at this stage.
Akim> But both CVS and 1.4 refuse sbin_DATA :(
Sorry, ba
Tom Tromey <[EMAIL PROTECTED]> writes:
> > "Akim" == akim <[EMAIL PROTECTED]> writes:
>
> Akim> My bad, I forgot to remove `dir' from the _DATA var.
>
> Oops, I missed that too.
>
> Akim> mybindir = $(bindir)/my
> Akim> mybin_DATA = doto
>
> User defined variables are handled by name.
>
> "Akim" == akim <[EMAIL PROTECTED]> writes:
Akim> My bad, I forgot to remove `dir' from the _DATA var.
Oops, I missed that too.
Akim> mybindir = $(bindir)/my
Akim> mybin_DATA = doto
User defined variables are handled by name.
I agree that this case is handled correctly with the current c
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> .am:19: invalid variable `mysysconfdir_DATA'
This should read `Makefile.am:19:...'.
Do you know why it broke?
Tom
On Sun, Feb 25, 2001 at 06:00:29PM -0700, Tom Tromey wrote:
>
> Akim> mysbindir = $(sbin)/my
> Akim> mysbindir_DATA = data
>
> Akim> .am:11: invalid variable `mysbindir_DATA'
>
> This is a new failure, since this feature has worked for a long time.
My bad, I forgot to remove `dir' from the _DA
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> I think I have this right too, since not all the *.am file
Akim> depend on ?EXEC?. See SCRIPTS and PROGRAMS below.
The point is that it is the directory name and not the primary that
determines which install-* target to use.
For i
Tom Tromey <[EMAIL PROTECTED]> writes:
> > "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
>
> Akim> . &transform_cond ('BASE' => $strip_subdir,
> Akim> 'EXEC' => $X =~ /exec/)
>
> This is still wrong.
>
> U
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> . &transform_cond ('BASE' => $strip_subdir,
Akim>'EXEC' => $X =~ /exec/)
This is still wrong.
User-defined directory names follow the `exec'-name test.
autoconf/au
Tom Tromey <[EMAIL PROTECTED]> writes:
> One example of how the new code can fail is a system-dependent header
> file install:
>
> inclexecdir = $(exec_prefix)/include ## Just an example!
> inclexec_HEADERS = my-config.h
>
> With the new code the header will be installed by `install-d
On Sat, Feb 24, 2001 at 10:29:14AM -0700, Tom Tromey wrote:
> > "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
>
> Akim> Here is the updated version of the patch.
>
> This patch still has a problem. See below.
>
> Akim> Note that there is a new side effect to my patch: the
> Akim> (un
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> Here is the updated version of the patch.
This patch still has a problem. See below.
Akim> Note that there is a new side effect to my patch: the
Akim> (un)?install-man targets appeared (since you said they were
Akim> required) in
> "Tom" == Tom Tromey <[EMAIL PROTECTED]> writes:
Tom> 2. Pushing the hook out means that if the user has a -local as
Tom> well as a -hook then we have to rewrite the -local target to add
Tom> a new dependency. We've never rewritten user targets, so this is
Tom> a major(-ish) change to conte
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
>> If it is a dependency then it won't work properly with a parallel
>> make. That's because the -hook for must be run after all the other
>> rules.
Akim> Why don't we use dependencies to say that -hook must be run
Akim> last, instead of
Tom Tromey <[EMAIL PROTECTED]> writes:
> Akim> Below are first presented the Makefiles of Automake. There is
> Akim> obviously a problem, which I do not solve in this already too
> Akim> long patch, but will address later: what do with empty targets.
>
> This patch is ok, but please don't check
Tom Tromey <[EMAIL PROTECTED]> writes:
> Targets which are presented to the user must be defined regardless of
> whether they have any effect. So for instance `install-exec' must
> always exist.
I understand this. Then I'll have a list of required targets and
output them in any case.
> Akim>
This patch applies the same transformation, factoring, to
install-data, install-exec and uninstall.
Below are first presented the Makefiles of Automake. There is
obviously a problem, which I do not solve in this already too long
patch, but will address later: what do with empty targets. For
ins
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> This patch applies the same transformation, factoring, to
Akim> install-data, install-exec and uninstall.
I haven't read this patch yet but I wanted to respond to some comments
before I did.
Akim> Below are first presented the Make
> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> This patch applies the same transformation, factoring, to
Akim> install-data, install-exec and uninstall.
Akim> Below are first presented the Makefiles of Automake. There is
Akim> obviously a problem, which I do not solve in this a
25 matches
Mail list logo
