Re: subjects not being parsed?

[Andrew Gray]

Hi Chad and Milan,

>     CCW> OK.  I'm at a loss here.  A couple months ago, I deployed gnats
>     CCW> 3.999 packages from the woody distribution of Debian.  During
>     CCW> that deployment and testing phase, I recall file-pr correctly
>     CCW> parsing the subject line and adding audit text to the PR.

> Yes, there was change in subject matching. The matching was changed,
> to improve a correct recognition of new reports and replies. It's
> not unlikely bugs were made during the change.
> 
> If you can, please look at the function `checkIfReply' in file-pr.c.
> Maybe you'll find out what's wrong.

I saw a similar problem with email's that were replies to existing PRs
not being recognised as such. I have had a look at the checkIfReply
function in file-pr.c and identified a couple of problems. The regular
expression currently used in checkIfReply is:

"\\<((PR[ \t/])|([-a-z0-9_+.]+)/)([0-9]+)"

(This is how it appears in the C source code, the \\ is
interpretted by the C compiler as a single backslash and \t as a
horizontal tab.)

Since the RE_NO_BK_VBAR bit defined in regex.h is not set using
re_set_syntax, \| is the alternation operator, so the vertical bar
should be preceded by a backslash in the regular expression:

"\\<((PR[ \t/])\\|([-a-z0-9_+.]+)/)([0-9]+)"

Currently checkIfReply uses re_match, but this function only tries to
match the regular expression at the start of the string. To match
subject lines that have miscellaneous text, such as "Re: " precedding
the PR identifier, the re_search function should be used.

Using the modified regular expression and re_search will match Subject
lines such as the following:

gnats/85
Re: gnats/85
FW: RE: gnats/85

Here is the patchfile to make these changes:

Index: file-pr.c
===================================================================
RCS file: /cvsroot/gnats/gnats/gnats/file-pr.c,v
retrieving revision 1.45
diff -u -p -r1.45 file-pr.c
--- file-pr.c   10 Feb 2002 18:23:42 -0000      1.45
+++ file-pr.c   7 Mar 2002 11:33:53 -0000
@@ -594,10 +594,11 @@ checkIfReply (PR *pr, ErrorDesc *err)
   regex.translate = case_fold;
   
   {
-    const char *const PAT = "\\<((PR[ \t/])|([-a-z0-9_+.]+)/)([0-9]+)";
+    const char *const PAT = "\\<((PR[ \t/])\\|([-a-z0-9_+.]+)/)([0-9]+)";
     re_compile_pattern (PAT, strlen (PAT), &regex);
   }
-  i = re_match (&regex, headerValue, strlen (headerValue), 0, &regs);
+  i = re_search (&regex, headerValue, strlen (headerValue), 0,
+                strlen (headerValue), &regs);
   regex.translate = NULL;
   regfree (&regex);
   re_set_syntax (old_syntax);

Regards,

Andrew Gray