On Wed, Aug 06, 2014 at 07:18:49PM +0200, Branko Čibej wrote: > On 06.08.2014 18:47, Bert Huijben wrote: > > Some time ago I switched a few of the calls mentioned here around to > > get a huge performance improvement on Windows. The standard integer > > parse functions start by skipping whitespace, and only then parse the > > integer. But whatever is whitespace is locale dependent, and checking > > for that from the CTY is really slow on Windows. > > (If I remember correctly these are exactly the parse id functions you > > are about to change here) > > In that case, these functions must still perform sanity checks.
If we can't use strtol for this purpose on Windows, then I think we should introduce our own replacement only on Windows. And also make sure our replacement is not any less correct in terms of how strtol() behaves otherwise.
