[Patch] Make add_item smarter
2004-05-30 Pierre Humblet <[EMAIL PROTECTED]>
* path.cc (mount_info::add_item): Make sure native path has drive
or UNC form. Call normalize_xxx_path instead of [back]slashify.
Remove test for double slashes. Reorganize to always debug_print. Index: path.cc
===
RCS file: /cvs/src/src/winsup/cygwin/path.cc,v
retrieving revision 1.313
diff -u -p -b -r1.313 path.cc
--- path.cc 28 May 2004 19:50:06 - 1.313
+++ path.cc 30 May 2004 04:06:43 -
@@ -2176,41 +2176,40 @@ mount_info::sort ()
int
mount_info::add_item (const char *native, const char *posix, unsigned mountflags, int
reg_p)
{
+ char nativetmp[CYG_MAX_PATH];
+ char posixtmp[CYG_MAX_PATH];
+ char *tail;
+ int err[2];
+
/* Something's wrong if either path is NULL or empty, or if it's
not a UNC or absolute path. */
- if ((native == NULL) || (*native == 0) ||
- (posix == NULL) || (*posix == 0) ||
- !isabspath (native) || !isabspath (posix) ||
+ if (native == NULL || *native == 0 || !isabspath (native) ||
+ !(is_unc_share (native) || isdrive (native)))
+err[0] = EINVAL;
+ else
+err[0] = normalize_win32_path (native, nativetmp, &tail);
+
+ if (posix == NULL || *posix == 0 || !isabspath (posix) ||
is_unc_share (posix) || isdrive (posix))
+err[1] = EINVAL;
+ else
+err[1] = normalize_posix_path (posix, posixtmp, &tail);
+
+ debug_printf ("%s[%s], %s[%s], %p",
+native, err[0]?"error":nativetmp, posix, err[1]?"error":posixtmp,
+mountflags);
+
+ if (err[0] || err[1])
{
- set_errno (EINVAL);
+ set_errno (err[0]?:err[1]);
return -1;
}
/* Make sure both paths do not end in /. */
- char nativetmp[CYG_MAX_PATH];
- char posixtmp[CYG_MAX_PATH];
-
- backslashify (native, nativetmp, 0);
nofinalslash (nativetmp, nativetmp);
-
- slashify (posix, posixtmp, 0);
nofinalslash (posixtmp, posixtmp);
- debug_printf ("%s[%s], %s[%s], %p",
- native, nativetmp, posix, posixtmp, mountflags);
-
- /* Duplicate /'s in path are an error. */
- for (char *p = posixtmp + 1; *p; ++p)
-{
- if (p[-1] == '/' && p[0] == '/')
- {
- set_errno (EINVAL);
- return -1;
- }
-}
-
/* Write over an existing mount item with the same POSIX path if
it exists and is from the same registry area. */
int i;
Re: [Patch] Make add_item smarter
On Sun, May 30, 2004 at 12:21:48AM -0400, Pierre A. Humblet wrote:
>2004-05-30 Pierre Humblet <[EMAIL PROTECTED]>
>
> * path.cc (mount_info::add_item): Make sure native path has drive
> or UNC form. Call normalize_xxx_path instead of [back]slashify.
> Remove test for double slashes. Reorganize to always debug_print.
>Index: path.cc
>===
>RCS file: /cvs/src/src/winsup/cygwin/path.cc,v
>retrieving revision 1.313
>diff -u -p -b -r1.313 path.cc
>--- path.cc28 May 2004 19:50:06 - 1.313
>+++ path.cc30 May 2004 04:06:43 -
>@@ -2176,41 +2176,40 @@ mount_info::sort ()
> int
> mount_info::add_item (const char *native, const char *posix, unsigned mountflags,
> int reg_p)
> {
>+ char nativetmp[CYG_MAX_PATH];
>+ char posixtmp[CYG_MAX_PATH];
>+ char *tail;
>+ int err[2];
Why an array here? It's not really being used as an array.
> /* Something's wrong if either path is NULL or empty, or if it's
> not a UNC or absolute path. */
>
>- if ((native == NULL) || (*native == 0) ||
>- (posix == NULL) || (*posix == 0) ||
>- !isabspath (native) || !isabspath (posix) ||
>+ if (native == NULL || *native == 0 || !isabspath (native) ||
Do we need an "isabsdospath"? Checking for isabspath and isdrive is
a little redundant although the compiler would probably optimize
nicely.
>+ !(is_unc_share (native) || isdrive (native)))
>+err[0] = EINVAL;
>+ else
>+err[0] = normalize_win32_path (native, nativetmp, &tail);
>+
>+ if (posix == NULL || *posix == 0 || !isabspath (posix) ||
> is_unc_share (posix) || isdrive (posix))
>+err[1] = EINVAL;
>+ else
>+err[1] = normalize_posix_path (posix, posixtmp, &tail);
>+
>+ debug_printf ("%s[%s], %s[%s], %p",
>+native, err[0]?"error":nativetmp, posix, err[1]?"error":posixtmp,
>+mountflags);
>+
>+ if (err[0] || err[1])
> {
>- set_errno (EINVAL);
>+ set_errno (err[0]?:err[1]);
> return -1;
> }
>
> /* Make sure both paths do not end in /. */
>- char nativetmp[CYG_MAX_PATH];
>- char posixtmp[CYG_MAX_PATH];
>-
>- backslashify (native, nativetmp, 0);
> nofinalslash (nativetmp, nativetmp);
Do we still need the nofinalslash?
cgf
Re: [Patch] Make add_item smarter
Yes, we need to remove the final slash, it can be present at the output of normalize_path 150 761725 [main] mount 671605 mount_info::add_item: c://[c:\\], /hagfsfd/[/hagfsfd/], 0xA Pierre
Re: [Patch] Make add_item smarter
On Sun, May 30, 2004 at 12:48:13AM -0400, Pierre A. Humblet wrote: >Yes, we need to remove the final slash, it can be present at the >output of normalize_path > > 150 761725 [main] mount 671605 mount_info::add_item: c://[c:\\], >/hagfsfd/[/hagfsfd/], 0xA But wouldn't it be faster to just query *tail? cgf
Re: [Patch] Make add_item smarter
Yes, we could use tail, but then we need to add logic to preserve the first / . I was lazy, or perhaps speed is not so important here. Pierre
Re: ssp.c (usage): Add missing linefeed.
I went ahead and applied this. It doesn't appear to break anything. :) On Mon, 24 May 2004 05:11:51 +0100, John Paul Wallington <[EMAIL PROTECTED]> wrote: > > 2004-05-24 John Paul Wallington <[EMAIL PROTECTED]> > > * ssp.c (usage): Add missing linefeed. > > --- ssp.c 14 Feb 2004 19:43:07 + 1.8 > +++ ssp.c 24 May 2004 05:09:52 +0100 > @@ -801,7 +801,7 @@ usage (FILE * stream) > " ssp -v -s -l -d 0x61001000 0x6108 hello.exe\n" > "\n"); >if (stream == stderr) > -fprintf (stream, "Try '%s --help' for more information.", prog_name); > +fprintf (stream, "Try '%s --help' for more information.\n", prog_name); >exit (stream == stderr ? 1 : 0); > } >
Re: [UG Patch] kmem and check_case typo
> >On Fri, 21 May 2004 10:22:20 -0500, Brian Ford wrote: > >> Ok, then shouldn't we apply the following patch to the users > >> guide? (plus a typo fix) Applied with the "planned for development" euphamism.
