Making things work on Windows

[Therem Harth]

Mmm? I thought AnonymousHero's version of T2 didn't use the Makefile.* any more, only CMake?

[nppangband]

I missed the T2 part. I thought you were talking about Angband. (translation: I assumed you were talking about Angband without reading because I didn't read the thread carefully enough )

[zaimoni]

Originally Posted by Therem Harth

Yeah, it's MinGW.

I tried the following BTW:

- Add MinGW to PATH
- Run CMake to generate a makefile
- Open an msysgit terminal (i.e. git bash), cd to the relevant directory, and run make

... make just prints

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation.  All rights reserved.

and kills bash, exiting to the normal Windows command shell. Running it again continues to print that message, and do nothing else.

The lead developer of CMake, Bill Hoffman, explicitly refused my offer to patch CMake 2.6 to handle MingW32 + MingW32 bash. (The officially supported check for MingW32 bails if it sees bash in the path.) It's an intentionally broken configuration. The fix is moderately easy (a "trust me" configuration).

There is a standoff between the CMake dev team and the CygWin dev team over what CMake on CygWin should define. As near as I could follow the flamewar on the CMake mailing list between the head developers, CygWin intentionally broke backward compatibility of CygWin and expected CMake to match CygWin; this was explicitly refused.

[Therem Harth]

The clashes of giant egos... IOW, why Microsoft is still eating everyone's lunch.

[Magnate]

Originally Posted by Therem Harth

The clashes of giant egos... IOW, why Microsoft is still eating everyone's lunch.

+1. I find these kind of disputes really upsetting, especially when it's so clear that neither solution is technically superior to the other, they're just a matter of preference. The inability to be objective about what's best in the long run is really sad.

Did Bill Hoffman say why he wanted Cmake + mingw32 + mingw32-bash to be intentionally broken? The reasoning isn't obvious.

[Therem Harth]

I find it grimly funny, in this case, that the "C" in "CMake" stands for "cross platform."

[zaimoni]

Originally Posted by Magnate

+1. I find these kind of disputes really upsetting, especially when it's so clear that neither solution is technically superior to the other, they're just a matter of preference. The inability to be objective about what's best in the long run is really sad.

Did Bill Hoffman say why he wanted Cmake + mingw32 + mingw32-bash to be intentionally broken? The reasoning isn't obvious.

No, he did not. Not even after I explicitlly diagnosed why it was broken in the CMake bug tracker, and explicitly explained on the mailing list to another would-be user of CMake on MingW32 + MingW32-bash how to bypass the broken MingW32 configuration.

He just doesn't want it fixed, period. I think patches to fix would be explicitly rejected.

[zaimoni]

Originally Posted by Magnate

+1. I find these kind of disputes really upsetting, especially when it's so clear that neither solution is technically superior to the other, they're just a matter of preference.

The CygWin dev team decided that they wanted to be so *NIX that a program compiled under CygWin must use POSIX headers completely; mixing in Windows headers will break the build. This was a radical policy change from the initially officially requested policy which was to treat CygWin as a Windows variant.

Both partiies agreed that the requested change was impossible to maintain backward compatibility for. As lead developer for CMake, Bill Hoffman basically told the CygWin representative that backward compatibilty was more important than supporting current CygWin, and that was why the year-old patch to implement the CygWin-desired change had been rejected. The counter-proposal to use an option to allow the CygWin CMake port to diverge from upstream's insistence on backward compability, was not publicly accepted.