Lars Gullik Bjønnes wrote:
> Angus Leeming wrote:
> | Rob Bearman wrote:
>>> With the recent changes to forkedcontr.h and forkedcall.*, I found that
>>> I needed to reinstate the typedef shown below to forkedcall.h and
>>> forkedcontr.h:
>>>
>>> #ifdef _WIN32
>>> typedef int pid_t;
>>> #endif
>>
> | I think that this one is fine, although it should become
>
> Not possible to put this in os_win32.h?
We want to get rid of os_win32.h. It's just a bunch of evil kludges.
> or make configure check for it and set it in config.h?
I guess that it is *possible*, but really do we care? How would user code
look?
PID_TYPE foo();
or would config.h have the typedef?
>>> I also needed to reinstate the #include "os_win32.h" to forkedcall.C,
>>> which I think is necessary to handle min/max naming conflicts.
>>
> | Well, given that the idea is eventually to remove this file entirely,
> | perhaps you'll first see whether this cures the problem:
>>
> | - int const tolerance = std::max(0, tol);
> | + int const tolerance = (std::max)(0, tol);
>
> Is this the same trick used on boost?
Yes. They went for a BOOST_STD_MAX or something for a while, but found that
it didn't work very well. I copied the above from some boost/python header
file.
--
Angus
