[PATCH] Regexp for RCScan

[bdbaddog]

This issue was originally created at: 2010-02-04 08:32:16.
This issue was reported by: gerard_.
gerard_ said at 2010-02-04 08:32:16

This patch (a new regexp) solves two problems:

1. It allows for custom resource types (like XAML). (this is an enhancement)
2. It does match when there are \r\n line endings (this is a bugfix)

gerard_ said at 2010-02-04 08:32:46

Created an attachment (id=686)
Regexp for RCScan

gerard_ said at 2010-02-04 10:03:36

Set type to PATCH

gregnoel said at 2010-02-19 11:14:09

Bug party triage. Steven to get advice from the .rc expert at his day job.

stevenknight said at 2010-02-22 15:58:37

I'm studying this issue, and wanted to capture an update with links to the
relevant MSDN spec pages for user-defined resources:

http://msdn.microsoft.com/en-us/library/aa380599(VS.85).aspx
http://msdn.microsoft.com/en-us/library/aa381043(VS.85).aspx
http://msdn.microsoft.com/en-us/library/aa381054(VS.85).aspx

There are two key questions:

1. the new regex is more expansive in matching any user-defined resource, and
   it's not clear whether we'd get false positives on lines that match two keywords
   followed by a quoted string that isn't a file name.

2. the old regex would match non-quoted file names following the specific
   keywords involved, while the new regex requires quotes. The spec pages I
   looked at don't indicate whether a file name must be quoted. If they can be
   left un-quoted, then we need to preserve the ability to match those.

gregnoel said at 2010-07-21 17:00:30

Bug party triage. Bump the priority of this issue.

gerard_ said at 2011-08-17 01:52:18

Steven, thanks for the links, on your questions:

1. The specification is indeed very broad. I think it's this:
   r'^\s*'+ # leading whitespace
   r'(?:\w+\s+\w+)'+ # BLAH BLAH
   r'\s+("[\w./\]+"|[\w./\]+)\s*$' # ..\file_haha.test or "file.test"
   (only for resources, not #includes, not tested)
   It would be much better to have two separate regexes instead of the one that
   tries to match both the resource and the #include.

If we want to prevent false positives it should check if it is a valid resource
type, and allow the user to provide user defined types.

2. I agree, it's a messy specification. See the regex from 1.

gerard_ attached RCScanner.patch at 2010-02-04 08:32:46.

Regexp for RCScan