Patch to make Windows build use autotools

[Daniel Atallah]

On Wed, Aug 25, 2010 at 00:20, John Bailey <rekkanoryo at rekkanoryo.org> wrote:
> On 08/24/2010 02:04 PM, Daniel Atallah wrote:
>> Folks,
>>
>> A couple months ago, a patch was submitted to make the autotools stuff
>> capable of building Pidgin (http://developer.pidgin.im/ticket/12218).
>>
>> I think it is a reasonable thing to do for uses like that ticket
>> mentions, but I don't have any plans for this to replace the existing
>> hand edited mingw Makefiles (mainly because it is a PITA to set up a
>> working autotools build environment on Windows).
>>
>> I'd like to get the thoughts and opinions of other folks about committing this.
>
> My primary concern is that this is going to cause even more confusion than there
> already is about Windows building.  As things stand now, when the build blows up
> and someone asks about it, we have the simple answer of "we don't support using
> the configure script on Windows."  This is particularly helpful currently
> because only a handful of us that are ever active in #pidgin and on the support
> list have a clue how to construct a Windows build environment.  If the configure
> script becomes a "viable" option for people, they're going to come seeking
> support when they can't figure out what they screwed up and even fewer people
> among us are going to be able to help.  (I likely wouldn't get up to speed on it
> anytime soon, and I can't imagine that there will be a mad rush on it.)

This is also my major concern about a change like this.

For the foreseeable future, it would remain an "unsupported" way to
build stuff on Windows and would quite possibly be broken without
anyone noticing for periods of time.

> A lesser, but still valid, concern is perhaps something that most people aren't
> going to care about, but I'll throw it out there anyway.  If I'm building
> something with cygwin and I use the configure script, I expect that it will
> build the package *for* cygwin.  Currently that doesn't work with Pidgin.  If we
> advertise that the configure script is now usable on Windows, we're going to
> have at least some people thinking that means they should be able to build
> cygwin binaries, when that won't be the case.

I don't know what the current status is of building a Cygwin version
of Pidgin.  I imagine that it wouldn't take much effort to do so, but
it is very much a "Patches Welcome" category from my perspective.  I
have no intention of making any effort to figure out what it would
take.

This patch would likely make it easier if someone wanted to do it, but
I really can't imagine a good reason for doing so.

I'm not worried about cygwin at all - anyone wanting to build a gui
application for cygwin should already have a reasonable expectation
that it'll take some effort.

> Overall, while I agree that on the surface this seems reasonable for the use
> cases described on the ticket, I'm not so sure I like the idea of doing it.  If
> this were a 100% foolproof way of doing things and it were reasonable to use
> this and get rid of the hand-crafted Makefile.mingw's all over the tree, I'd be
> much more inclined to agree with it.

It is possible that this would eventually become the supported way to
build Pidgin/libpurple on Windows eventually, but I really don't want
to consider that right now.

My opinion is that as long as this isn't going to break or
significantly complicate the existing autotools stuff for other
platforms, it should be accepted with the caveat that it is not the
"officially supported" (whatever that really means) way of building
for Windows and that maintenance is considered to be on a "Patches
Welcome" basis in the same way that the implementation came about.
Documentation to that effect could be added to the wiki.

-D