Re: Possible eval builtin speedup?

[Chet Ramey]

Nicolas wrote:
> Jan Schampera wrote:
> > Yea, I didn't look close enough. My fault.
>
> Ok, I found the problem: xdupmbstowcs (in file /lib/glob/xmbsrtowcs.c)
> is painfully slow, because of very frequent calls to realloc. I think
> this could be optimized.

You're probably right, and I see at least one way that could be done.
I will happily accept any contributions to speed it up.

> Nevertheless, I don't really understand the purpose of this function.
> Is it supposed to bypass a bug in mbsrtowcs (defined in wchar.h), as
> said in the following commentaries?
> /* xmbsrtowcs.c -- replacement function for mbsrtowcs */
> /* On some locales (ex. ja_JP.sjis), mbsrtowc doesn't convert 0x5c to U<0x5c>.
>    So, this function is made for converting 0x5c to U<0x5c>. */
> If so, why isn't it corrected to wchar.h?

I don't really understand this question.

Chet
--
``The lyf so short, the craft so long to lerne.'' - Chaucer
                       Live Strong.  No day but today.
Chet Ramey, ITS, CWRU    chet@case.edu    http://cnswww.cns.cwru.edu/~chet/