[Top][All Lists]

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

Re: [Mingw-users] MSYS_PREFIX_HACK

From: Keith MARSHALL
Subject: Re: [Mingw-users] MSYS_PREFIX_HACK
Date: Fri, 1 Apr 2005 11:18:29 +0100

Earnie Boyd wrote:
>> 1) Variable `ac_default_prefix' is explicitly initialised,
>>    as `ac_default_prefix=/usr/local'
>> 2) Variable `prefix' is explicitly initialised, as
>>    `prefix=NONE'
> Relative paths are not converted so the value of prefix will be 
>> 3) Command line arguments are parsed; if `--prefix=value'
>>    is specified, then variable `prefix' is redefined, as
>>    `prefix=value'
>> Then, when we get to the AC_OUTPUT phase, I see:
>>   `test "x$prefix" = xNONE && prefix=$ac_default_prefix'
> So when prefix is set to /usr/local it is converted to the win32 path
> (E.G.: C:/MSYS/1.0/local).  I've used my change enough now to know that 
> works well.  I want to do a few more configure exercises before I upload 
> snapshot.

That seems pretty much what we want, for the purposes of configuring and
making autotool managed packages.  However, still playing devil's 
my concern is not that this won't work -- I'm sure you've made it work
very well indeed -- but that it is making MSYS, and its implementation
of `bash', unnecessarily specialised towards this goal.

I appreciate that this *is* the primary goal of MSYS; however, it is also
genuinely useful far beyond this primary goal.  Agreed, attaching special
significance to an environment variable called `prefix' is a small price
to pay, in order to better realise the goal; I do wonder, however, if it
is the best solution to a problem which could be eliminated at source, by
a fairly trivial patch to autoconf's `general.m4' code, to make it more
intelligent in its initial assignment of `ac_default_prefix'.

As a short term solution, this MSYS change sounds ideal; I look forward
to trying it out, when the snapshot is available.  In the long term, I
still think it is worth pursuing the idea of an autoconf patch.  When
that eventually filtered through to the majority of packages, the need
for specialised handling of `prefix' would fall away.  Perhaps it would
never become completely unnecessary however; how feasible would it be to
provide this special `prefix' handling feature, but to also incorporate
some mechanism, say with a custom `shopt' or `set -o' option, to enable
and disable it at will?

> If you care to suggest a package configure to try, feel free to let
> me know about it.

The one with which I am most closely involved is `groff' -- the current
CVS, from, builds OOTB when `prefix' is correctly
defined, although there are some additional dependencies beyond MSYS
and MinGW, which I have satisfied with packages from the GnuWin32
project,, and GhostScript

Best regards,

reply via email to

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