[tex-k] New tl build process (fwd)

Taco Hoekwater taco at elvenkind.com
Wed Mar 4 12:49:57 CET 2009


Peter Breitenlohner wrote:

> There are, however, a few minor points and a major one.
> 
> In mingw32.h IS_UNC_NAME is defined twice (formatted differently).

I'll fix that.

> In config.h there now is
>     #ifdef WIN32
>     #ifndef __STDC__
>     #define __STDC__ 1
>     #endif
>     #endif /* not WIN32 */
> whereas I'd (mildly) prefer
>     #ifdef WIN32
>     #undef __STDC__
>     #define __STDC__ 1
>     #endif /* not WIN32 */

But that won't rid the terminal of gcc warning messages as it will
just trigger a different one.

> In tex-make.c, instead of introducing a new variable NUM, I'd rather define
> num as DWORD for mingw and as int otherwise (BTW: what is the 
> difference, or
> in other words is mingw's int 16Bit?, or maybe signedness?).

DWORD comes from Microsoft (as is ReadFile), and is an unsigned long.

  BOOL WINAPI ReadFile(
   __in         HANDLE hFile,
   __out        LPVOID lpBuffer,
   __in         DWORD nNumberOfBytesToRead,
   __out_opt    LPDWORD lpNumberOfBytesRead,
   __inout_opt  LPOVERLAPPED lpOverlapped
);

(LPDWORD is a DWORD *)


> Still in tex-make.c: Taco, you can't just unconditionally add a third
> argument to look_for_cmd, that will break non-MINGW builds.  Moreover
> the use of that third argument `&new_cmd' clashes with the current code:
> immediately after a successful call to look_for_cmd a new value is
> assigned to new_cmd.

So where is the two-argument version of look_for_cmd defined? Is that
in some file Akira has?

> I think it would be best if native WIN32 and MINGW had the same API for
> that function

Agreed, but the native version is not in the repository and
I inherited the three-argument version from Siep.

Best wishes,
Taco


More information about the tex-k mailing list