On Thu, Apr 09, 2020 at 05:00:34PM +0200, Martin Ågren wrote:
> > That explains why I didn't see the issue when running doc-diff on the
> > original bug. I wonder if we could teach doc-diff to look at the HTML,
> > too. I'm not sure how, though. Certainly html2text or similar would get
> > us something diff-able, but without the visual elements (like the grey
> > box), I don't know that it's much more valuable than the manpages.
>
> At one point I considered trying out diffoscope for this. It should
> allegedly be good at comparing "everything". But being good at
> everything, it wanted to pull in a discouragingly large number of
> dependencies, so I never actually tried it out. It doesn't explicitly
> claim to know html or manpages (but does mention xml and pdf), so I
> dunno.
I tried it just now, and it's not that clever. A regular "diff -r" of
the before and after HTML yields what you'd expect:
--- old/git-config.html 2020-04-09 11:38:19.312436125 -0400
+++ new/git-config.html 2020-04-09 11:38:40.028385850 -0400
@@ -1678,11 +1678,9 @@
; file (if the condition is true); their location is not
; affected by the condition
[includeIf "gitdir:/path/to/group/"]
- path = foo.inc</code></pre>
-</div></div>
-<div class="literalblock">
-<div class="content">
-<pre><code>; include only if we are in a worktree where foo-branch is
+ path = foo.inc
+
+; include only if we are in a worktree where foo-branch is
; currently checked out
[includeIf "onbranch:foo-branch"]
path = foo.inc</code></pre>
A diffoscope diff yields the same, plus it complains about differing
timestamps on all of the files. I don't think it's doing anything clever
with respect to HTML formatting.
-Peff
