Re: [PATCH] fix problem with uconvert

[Isabella Bosia]

Dunno if it's "considerably easier", my patch just added 1 sign argument to
1 function and fixed up a couple of call sites.
Your way is shorter but it boils down to "if the micro-part is negative it
means that the number was in (-1,0)", which for some reason was already
being checked for despite not being possible before.
But regardless, I only care for this to be fixed, not how.
Cheers,
iza

On Mon, 24 Oct 2022, 19:54 Chet Ramey, <chet.ramey@case.edu> wrote:

> On 10/22/22 2:58 PM, izabera wrote:
> > if the argument is in the (-1, 0) range, the integer part is zero and
> > multiplying it by -1 has no effect, so the caller can't tell that the
> > argument was negative
>
> I suggest that a considerably easier way to solve this is
>
> *** ../bash-5.2-patched/lib/sh/uconvert.c       2020-06-04
> 12:28:58.000000000 -0400
> --- lib/sh/uconvert.c   2022-10-24 14:50:15.000000000 -0400
> ***************
> *** 41,45 ****
>    do { \
>      if (ip) *ip = ipart * mult; \
> !   if (up) *up = upart; \
>      if (ep) *ep = p; \
>      return (x); \
> --- 41,45 ----
>    do { \
>      if (ip) *ip = ipart * mult; \
> !   if (up) *up = upart * (ipart == 0 ? mult : 1); \
>      if (ep) *ep = p; \
>      return (x); \
>
>
> --
> ``The lyf so short, the craft so long to lerne.'' - Chaucer
>                  ``Ars longa, vita brevis'' - Hippocrates
> Chet Ramey, UTech, CWRU    chet@case.edu    http://tiswww.cwru.edu/~chet/
>
>