bugzillas+padREMOVETHISdu wrote:

Here's another URL that shows the bug:

The actual code produced is:
<tr><td>-</td><td class="diff-deletedline">This system of
measurement also introduces separators into numbers in a different
place than is common outside India. For example, 3 million (30 lakh)
would be written as 30,00,000.</td><td>+</td><td class="diff-addedline">This
system of measurement also introduces separators into numbers in a
different place than is common outside India. For example, 3 million
(30 lakh) would be written as 30,000,000.</td></tr>

The expected code is:
<tr><td>-</td><td class="diff-deletedline">This system of
measurement also introduces separators into numbers in a different
place than is common outside India. For example, 3 million (30 lakh)
would be written as 30,00,000.</td><td>+</td><td class="diff-addedline">This
system of measurement also introduces separators into numbers in a
different place than is common outside India. For example, 3 million
(30 lakh) would be written as 30,00<span class="diffchange">0</span>,000.</td></tr>

				   ^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^

bugzillas+padREMOVETHISdu wrote:

(In reply to comment #1)

Here's another URL that shows the bug:

The URL is (as in the URL field above):
http://en.wikipedia.org/w/wiki.phtml?title=Lakh&diff=6163130&oldid=6115828

alarm wrote:

a small change (!= into !== to treat '01' different from '001')

attachment DifferenceEngine.bug349patch.txt ignored as obsolete

alarm wrote:

The patch #3 does not solve initial issue, but solves #2 issue (with digits treated as numbers, not as text

Committed patch for second issue to CVS HEAD.

gangleri wrote:

(In reply to comment #5)

Committed patch for second issue to CVS HEAD.

The patch fixed
bug 2824 diffs and literal compare of numbers fails

Thanks! best regards reinhardt [[user:gangleri]]

ezyang wrote:

Here's another testcase for the bug:

http://en.wikipedia.org/w/index.php?title=FanFiction.Net&curid=1530445&diff=32031545&oldid=31590107

Quite annoying actually. I bet it has something to do with control
characters/whitespace.

gangleri wrote:

(In reply to comment #1)

http://en.wikipedia.org/w/index.php?title=Lakh&action=edit&oldid=6163130
... 30,000,000.
http://en.wikipedia.org/w/index.php?title=Lakh&action=edit&oldid=6115828
... 30,00,000.
is *not* covered by attachment.cgi id=1190

Here are more numbers separated with "," other countries are using ".". Please
change the code that diff sees three "number strings" . Compare only sequences
of characters between 0 and 9, do not include ",", ".", "e" or "E" as in 10E3.

Please note that only the testcases from
bug 2824: diffs and literal compare of numbers fails
are fixed. I will reopen bug 2824 because "30,000,000" is a valid number.

(In reply to comment #7)
old
http://en.wikipedia.org/w/index.php?title=FanFiction.Net&action=edit&oldid=31590107
'''FanFiction.Net''' ...
new
http://en.wikipedia.org/w/index.php?title=FanFiction.Net&action=edit&oldid=32031545
'''FanFiction. Net''' ...

This was *no* whitespace. It is a "normal" space. You will not find any
"unusual" Unicode character at
https://bugzilla.mozilla.org/show_bug.cgi?id=318151#c7 .

adding depends on:
bug 3894: space, white space characters, BiDi control characters should show up
in diff

It is hard to find the original "permanent link". Tray at
http://en.wikipedia.org/w/index.php?title=Western_philosophy&diff=next&oldid=813738

best regards reinhardt [[user:gangleri]]

gangleri wrote:

changed dependencies
the actual bug depends on 2824

gangleri wrote:

screen shot for bug 349

url:
http://yi.wiktionary.org/w/index.php?title=project:bugzilla/unicode/U_00A0&diff=7720&oldid=7186

The "&" in '%c2&a0' was changed to "%".
It is *not* red in the older revision but it *is* red in the newer.

*note*
The screen shot was made with [[wikt:yi:user:Gangleri/monobook.js]] where
[[en:Wikipedia:WikiProject User scripts/Scripts/Fix diff width]] from
Bug 1229: Balance diff display more evenly
is used.

best regards reinhardt [[user:gangleri]]

Attached:

gangleri wrote:

Don't know if this belongs here:
http://yi.wiktionary.org/w/index.php?title=project:bugzilla/unicode/U_00A0&diff=7733&oldid=7720
Diff highlights "to much". At the url a long line was splitted into more lines
in order to improuve readability.

Do not use anything on Wikimedia sites at this time for testing the internal
diff engine, as we're using an external diff engine now (wikidiff2).

bugzillas+padREMOVETHISdu wrote:

Why was this marked invalid? Either the original issue is fixed or it is not.

BTW
http://yi.wiktionary.org/w/index.php?title=project:bugzilla/unicode/U_00A0&diff=7720&oldid=7186
is still showing the issue even for an anonymous account ('%c2&a0' not
highlighted in the older revision (to the right)) though I'm not sure if this
should be discussed here or in a new bug number.

fentonb wrote:

I don't know if this is the same issue, but sometimes when doing a Diff in en.wikipedia.org, text that hasn't actually been deleted gets marked in red in the diff. This happened to me today when I mistakenly reverted this edit (http://en.wikipedia.org/w/index.php?title=Sigur_R%C3%B3s&diff=145598482&oldid=145347392 ). I thought the editor had deleted text when he hadn't. So to me, it looks like a bug in diff.

Should I post a new bug or is this comment sufficient?

thanks
Brian

alexsm333 wrote:

#14 Brian Fenton: diff works by paragraphs. From Mediawiki point of view, that text was removed from the 1st paragraph and then inserted later. This example doesn't look like a bug to me, simply a shortcoming of the current diff system.

fentonb wrote:

(In reply to comment #15)
AlexSM: thanks for replying. Yes I understand that, but there was no new paragraph. All the editor did was add a ref tag in the middle of the paragraph. The ref should have been highlighted as the change.

Changed example URL to http://en.wikipedia.org/w/wiki.phtml?title=Philosophy&diff=next&oldid=813738 which shows the original problem (paragraph listed as a remove + add instead of changed with individual words changed). This is still present in the wikidiff2 version.

The number issue which the attached patch fixes was separately reported at bug 2824 (now marked fixed, since the patch fixed it).

Paragraph splitting issue mentioned in the couple comments above is bug 5072.

*Bulk BZ Change: +Patch to open bugs with patches attached that are missing the keyword*

[[User:Cacycle/wikEdDiff]] does diffs in a non-paragraph manner.

sumanah wrote:

John: you marked the patch obsolete because it doesn't cleanly apply against current trunk, right?

Is the issue in its current form still happening in 1.17.0?

Per comment 17, a problem with line matching.

• This bug has been marked as a duplicate of bug 13462 ***