SEO

#1 (**permalink**) 02-18-2008, 07:22 AM

On Wed, 27 Oct 2004, Jared Yanovich wrote:

> On Tue, 26 Oct 2004 08:54:13 +0200 (CEST)
> Otto Moerbeek <otto@drijf.net> wrote:
>
> > > > - cvs also uses a slightly different date string plus a "<tab>revision"
> > >
> > > Yes, but patch parses the filename from the "Index:" line from
> > > cvs-generated diffs, so I don't think this matters.
> >
> > The Index: line is optional, and thins are also dependent if filenames
> > mentioned do exist or not. Check
> > <http://www.opengroup.org/onlinepubs/000095399/utilities/patch.html> (you
> > are wrong if you think standards are there to make life simple)
>
> Okay
>
> > I have this diff in my tree, which is supposed to handle any funny
> > character in pathnames _except_ tabs. I was not able to find a simple,
> > robust way to handle tabs.
>
> The approach I took made some assumptions that won't really work, now
> that I think about it. I think your handling of tabs is the simplest
> solution so far, but I think any other space character should break the
> loop, not just a tab when one is present.
>
> So maybe this,
>
> Index: util.c
> ================================================== =================
> RCS file: /cvs/src/usr.bin/patch/util.c,v
> retrieving revision 1.28
> diff -u -p -r1.28 util.c
> --- util.c 5 Aug 2004 21:47:24 -0000 1.28
> +++ util.c 27 Oct 2004 04:18:20 -0000
> @@ -333,7 +333,7 @@ char *
> fetchname(const char *at, bool *exists, int strip_leading)
> {
> char *fullname, *name, *t;
> - int sleading;
> + int sleading, tab;
> struct stat filestat;
>
> if (at == NULL || *at == '\0')
> @@ -349,8 +349,10 @@ fetchname(const char *at, bool *exists,
> return NULL;
> name = fullname = t = savestr(at);
>
> + tab = strchr(t, '\t') != NULL;
> /* Strip off up to `strip_leading' path components and NUL terminate. */
> - for (sleading = strip_leading; *t != '\0' && !isspace(*t); t++) {
> + for (sleading = strip_leading; *t != '\0' && (!isspace(*t) ||
> + (tab && *t == ' ')); t++) {
> if (t[0] == '/' && t[1] != '/' && t[1] != '\0')
> if (--sleading >= 0)
> name = t + 1;
>
> It obviously fails on the following two cases, but I would like to
> think they are not very common.
>
> --- a really/long and/strange filename.txt Sun Oct 24 20:51:25 2004
> +++ t/b Sun Oct 24 20:51:32 2004
>
> --- a filename Sun Oct 24 20:51:25 2004
> +++ t/b Sun Oct 24 20:51:32 2004
>

If there is a space before the date string I follow this, otherwise this
just confuses me. My diff was designed to fall back to the old behavior
when no tabs are found (i.e. we have a diff that follows Posix strictly).
So far I have not seen a program that actually produces these. Anybody
with a Solaris machine care to check what diff does on those?

The question is if we want to handle non-space (as in != ' ') isspace()
chars. I agree that things like vertical tabs and form feed in pathnames
are really strange, so that might not be needed. But on the other hand, if
the only exception is tab, that's more easy to follow and document: if a
tab occurs in the line, it is used as a separator, otherwise any isspace()
char is used as a separator.

So I have a slight preference for my original diff.

-Otto

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

SEO

Re: patch(1) w/paths containing spaces