[Lldb-commits] [PATCH] Make build work on MinGW

[Zachary Turner]

I actually think a better solution is to just stop building PseudoTerminal
stuff on Windows.  That's a larger change, but I suspect I'll do it at some
point in the future.


On Thu, Jul 31, 2014 at 2:44 PM, Zachary Turner <zturner at google.com> wrote:

> On Thu, Jul 31, 2014 at 2:30 PM, Keno Fischer <
> kfischer at college.harvard.edu> wrote:
>
>>
>> ================
>> Comment at: source/Utility/PseudoTerminal.cpp:23-27
>> @@ -22,2 +22,7 @@
>> +
>> +#ifndef __MINGW32__
>> +#ifndef LLDB_DISABLE_PYTHON
>>  typedef uint32_t pid_t;
>> +#endif
>> +#endif
>>  // empty functions
>> ----------------
>> Zachary Turner wrote:
>> > Little confused about what's going on here and why we're messing with
>> pid_t in two different places (this file and win32.h) with different
>> conditions.  Is it possible to centralize all the logic about what's going
>> on with pid_t into win32.h?
>> Yes, I think that would be a good idea. I was trying to do the inverse
>> condition of the one in win32.h here, to avoid it being defined twice. I
>> think we should unify it in win32.h if possible.
>>
> I thought that might be what it was, but inverse would be #if
> !defined(__MINGW32__) || !defined(LLDB_DISABLE_PYTHON).  But then after you
> fix the original case, the inverse will be different.
>
> I thought about this some and here's what I think:  The correct condition
> in win32.h is #if defined(LLDB_DISABLE_PYTHON) && !defined(__MINGW32__)
>
> In PseudoTerminal.cpp I think you can use simply #if !defined(__MINGW32__)
>
> Note that the two aren't strict inverses.  If __MINGW32__ is defined, you
> never want pid_t because mingw headers already provide it.  So in neither
> case should it be defined.
>
> Python.h on native Win32 typedefs it separately though.  So in the header
> file, which is supposed to be general purpose and shouldn't conflict with
> Python.h, we only typedef pid_t if python is disabled since there's no
> chance of a conflict.  If Python is NOT disabled, then don't do anything,
> since it would make the two headers incompatible.
>
> In PseudoTerminal.cpp, we know we aren't ever going to include Python.h,
> so we can just always typedef it.
>
> It's kind of confusing though, so I hope I did that right.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-commits/attachments/20140731/34d3bc0d/attachment.html>