Bug 62965 – webkit-patch/svn-create-patch created patch with invalid chunk range lines

• Component: Tools / Tests
• Status: NEW

[Daniel Bates]

The attachment 97685 [details] (of bug #62613) failed to apply because it's malformed since it contains invalid chunk range lines.

For example, the diff for JSCParser.cpp contains the following chunk:

[[
@@ -71,8 +82,9 @@ COMPILE_ASSERT(LastUntaggedToken < 64, L
 
 class JSParser {
 public:
-    JSParser(Lexer*, JSGlobalData*, FunctionParameters*, bool isStrictContext, bool isFunction, SourceProvider*);
-    const char* parseProgram();
+    JSParser(Lexer*, JSGlobalData*, FunctionParameters*, bool isStrictContext, bool isFunction, const SourceCode*);
+    UString parseProgram();
 private:
     struct AllowInOverride {
         AllowInOverride(JSParser* parser)
]]

But the chunk range line should be @@ -71,8 +82,8 @@ (since we have 8 lines of text after the change - the same number of lines we started with)

[Daniel Bates]

Created attachment 97747 [details]
Malformed diff for JSCParser.cpp

Extracted from attachment 97685 [details].

[Daniel Bates]

Created attachment 97748 [details]
Reproduced diff for JSCParser.cpp (produced by svn diff)

Reproduced the JSCParser.cpp changes and then used svn diff to generate this diff.

[Oliver Hunt]

What the heck is svn-create-patch doing to break patches?  Why does it need to modify the diff at all?

/me cries at the horror that is svn

[Daniel Bates]

Created attachment 97750 [details]
Reproduced diff for JSCParser.cpp (produced by svn-create-patch)

Reproduced the JSCParser.cpp changes and then used svn-create-patch to generate this diff.

Notice that this diff is identical to the diff produced by svn diff (attachment 97748 [details]) up to the inlined context line to the right of the chunk range.

[Daniel Bates]

@Juan: Can you elaborate on how you created attachment 97685 [details]? Oliver Hunt stated that this patch was produced via webkit-patch (bug 62613 comment 32).

Assuming you still have this change applied to your working copy, can you run svn diff from the top-level WebKit directory, redirecting its output to a file, and post it as an attachment on this bug?

[Juan C. Montemayor]

(In reply to comment #5)
> @Juan: Can you elaborate on how you created attachment 97685 [details]? Oliver Hunt stated that this patch was produced via webkit-patch (bug 62613 comment 32).

Attachment 97685 [details] was created using svn-create-patch. I am currently creating a patch with webkit-patch to see if that works.  Additionally I will re-run svn-create-patch and diff that with the bad patch to see if anything has changed.

> Assuming you still have this change applied to your working copy, can you run svn diff from the top-level WebKit directory, redirecting its output to a file, and post it as an attachment on this bug?

Yep. Will attach diff in a minute.

[Juan C. Montemayor]

Created attachment 97820 [details]
requested diff.