On Fri, 4 Aug 2017, Fabrice Popineau wrote:

> 
> 
> 2017-08-04 16:49 GMT+02:00 Tino Calancha <tino.calancha@gmail.com>:
> 
>
>       On Fri, 4 Aug 2017, Fabrice Popineau wrote:
> 
> 
>
>             2017-08-04 16:23 GMT+02:00 Fabrice Popineau <fabrice.popineau@gmail.com>:
> 
>
>             Actually, the problem seems to be in the `insert-directory-wildcard-in-dir-p' function
>             which wrongly splits "c:/tmp/dir*/*.txt" in ("c:/tmp/" . "dir*/*.txt") instead of
>             ("c:/tmp/dir*/" . "*.txt")
>
>             Forget this (wrong) diagnostic.
>
>             The culprit is actually 
>
>             (let ((default-directory "c:/tmp/"))
>               (eshell-extended-glob "dir*/*.txt"))
>             "dir*/*.txt" 
>
>             which fails to expand the wildcards (when `file-expand-wildcards' succeeds).
>
>       Thank you Fabrice,
>       that's interesing.  I am just wondering if `eshell-extended-glob' gets confused with the Windows path, i mean, the
>       disk name 'c:' in front.
>
>       Could you check if the following works?
>       M-x eshell RET
>       cd "c:/tmp"
>       ls -l dir*/*.txt
> 
> It says : 
> 
> c:/tmp $ ls -l dir*/*.txt
> dir*/*.txt: No such file or directory 
I see.  I think hat uses `eshell-extended-glob' as well, so it seems
you find a bug in that function.   Maybe is better to open a bug
report with this eshell recipe above.
> 
>
>       I am also curious if:
>       M-: (equal temporary-file-directory "c:/tmp/") RET
>       => t
> 
> Nope, my temp dir in this case is the Windows temp dir.
> ("c:/Users/Fabrice/AppData/Roaming/Local/Temp/") 
> 
> And out of curiosity: what does em-glob that file-expand-wildcards doesn't? 
> Because replacing the former by the latter would withdraw a depency on eshell parts in ls-lisp :
I think you are right.  We must use `file-expand-wildcards'.
I think the reason i didn't use this func. is because i didn't know
it existance :-)