On Thu, Aug 03, 2017 at 04:13:38PM +0530, Prathamesh Kulkarni wrote:
> > You are hardcoding here host characters and using it for target.
> > I think you need to use
> > lang_hooks.to_target_charset
> > (really no idea how it works or doesn't in LTO, but gimple-fold.c is already
> > using it among others).
> > Also, you're assuming that the 'a'-'z' and 'A'-'Z' ranges are without gaps,
> > which isn't the case for e.g. EBCDIC. So I think you'd need to verify
> > (once?) that the target charset has this property.
> Hi Jakub,
> Thanks for the suggestions, I wasn't aware about lang_hooks.to_target_charset.
> Would following be a correct check that target has ascii charset and only then
> set range to ~[a, z] ?
>
> target_a = lang_hooks.to_target_charset ('a');
> target_z = lang_hooks.to_target_charset('z');
> if (target_a == 'a' && target_z == 'z')
> {
> // set vr to ~['a', 'z']
> }
No. E.g. if host == target charset is EBCDIC, then the condition is true,
but it isn't a consecutive range.
A rough check that would maybe work (at least would be false for EBCDIC
and true for ASCII) could be something like:
if (target_z > target_a && target_z - target_a == 25)
(which is not exact, some strange charset could have say '0'-'9' block in
the middle of 'a'-'z' block, and say 'h'-'r' outside of the range), but
perhaps good enough for real-world charsets.
In any case, you should probably investigate all the locales say in glibc or
some other big locale repository whether tolower/toupper have the expected
properties there.
Plus of course, the set vr to ~[ ] needs to use target_a/target_z.
Jakub
