Re: [PATCH] xheader: fix lookup of .-suffixed keyword

[alice]

On Tue Jul 18, 2023 at 5:21 PM UTC, Sergey Poznyakoff wrote:
> psykose <alice@ayaya.dev> ha escrit:
>
> > str[strlen(str)] returns the NUL byte, not the last character.
>
> That's irrelevant.  kwlen below is *not* the length of keyword, but
> that of p->keyword:
>
> >     size_t kwlen = strlen (p->keyword);
> > -   if (strncmp (p->keyword, keyword, kwlen) == 0 && keyword[kwlen] == '.')
> > +   if (strncmp (p->keyword, keyword, kwlen) == 0 && keyword[kwlen - 1] == 
> > '.')
>
> That patch actually breaks the functionality, instead of fixing it.

yeah, noticed it actually breaks the testsuite after...

the actual issue was that

--- a/src/xheader.c
+++ b/src/xheader.c
@@ -1777,5 +1777,8 @@ struct xhdr_tab const xhdr_tab[] = {
      they are restored *only once* during extraction later on. */
   { "SCHILY.xattr", xattr_coder, xattr_decoder, 0, true },
 
+  /* Ignore apk-tools headers */
+  { "APK-TOOLS.", dummy_coder, dummy_decoder, 0, true },
+
   { NULL, NULL, NULL, 0, false }
 };

(as a downstream patch)
used to work with 1.34 but no longer works with 1.35, and prints an unknown-
header warning.

i now realise that omitting the trailing `.` makes the strlen match up correctly
(since it will be one ahead, and land on the '.' of the header.) thanks for
looking! sorry for the noise.


>
> Regards,
> Sergey