Re: win32 short name and IFS='~'

[Guido Draheim]

Naofumi Yasufuku wrote:
> At Mon, 31 Mar 2003 00:10:10 +0200,
> Guido Draheim wrote:
>
> > I have a similar problem on a different account: the version
> > management system at my employer uses "~" in directory names
> > to flag different branches and subversions of projects and
> > their checkout areas. The libtool has the tendency to
> > resolve some symlinks, so it does not help to put some
> > directories elsewhere. It was impossible to build with
> > libtool in this environment - after some time I did write
> > up an ac-macro that changes the _cmds IFS from "~" into "?"
> > which is much more uncommon to exist in either a filename
> > or a _cmds specification. Add the following macro after
> > OUTPUT, and you should be fine with any "~" in file or
> > directory names in the system:
> >
> > http://ac-archive.sf.net/guidod/patch_libtool_changing_cmds_ifs.html
> >
> > Btw, I wouldn't mind if libtool would simply not use "~" as
> > the IFS in its original source, so this libtool-patching
> > could get obsolete once and for all. -- cheers, guido
>
>
> I think '~' IFS char is chosen carefully and it shouldn't be changed.
> For example, '?' might be used in regular expression, and so on.
> In libtool, directories are mostly treated as absolute path, so that
> '~' is relatively safe.

That's the point, I have a directory /my/project~tng/src and the
libtool will break at an unknown /my/project or tng/src. Where
do you think "?" would get in the way? makefile targets are
resolved anyway, and real-world file globs use "*" anyway, the
use of "?"-fileglobs looks academic too me - show me a counter
example, an existing one please.

-- cheers, guido