How to make `hg diff` show whitespace changes?

[Matt Mackall]

On Thu, 2016-01-14 at 11:04 -0500, Matt Harbison wrote:
> (cc: list)
> 
> On Thu, 14 Jan 2016 10:26:05 -0500, Noemi Millman  
> <noemi.millman at backstage.com> wrote:
> 
> > 
> > Thanks, Matt!
> > 
> > Great call about the resource forks (and a facepalm moment for me!)  I  
> > can confirm that that's the problem, since changing the default app to  
> > open a file without editing the file itself adds it to the list of  
> > mysteriously changed-yet-unchanged files.
> > 
> > I wonder why Mercurial recognizes resource fork changes but git doesn't?
> 
> It's been a few months since I looked into it, but IIRC, the python call  
> used to determine the file size sees both the data and resource forks.

First, let's all agree that BeyondCompare devs should not fare well in the
afterlife for adding resource forks to read-only files on Unix in 2015.

Everything I can find suggests that stat(2) and os.stat ignore the resource fork
for st_size (but include it in st_blocks, which we ignore). But we're actually
using the special getdirentriesattr interface:

https://www.selenic.com/hg/file/tip/mercurial/osutil.c#l483

And it's complicated. But it looks like instead of ATTR_FILE_TOTALSIZE, we
should be using ATTR_FILE_DATALENGTH:
https://developer.apple.com/library/ios/documentation/System/Conceptual/ManPages
_iPhoneOS/man2/getattrlist.2.html#//apple_ref/doc/man/2/getattrlist

Please test this patch:

diff -r 443848eece18 mercurial/osutil.c
--- a/mercurial/osutil.c	Wed Jan 13 15:47:37 2016 -0600
+++ b/mercurial/osutil.c	Thu Jan 14 12:20:28 2016 -0600
@@ -458,7 +458,7 @@
 	requested_attr.bitmapcount = ATTR_BIT_MAP_COUNT;
 	requested_attr.commonattr = (ATTR_CMN_NAME | ATTR_CMN_OBJTYPE |
 				     ATTR_CMN_MODTIME | ATTR_CMN_ACCESSMASK);
-	requested_attr.fileattr = ATTR_FILE_TOTALSIZE;
+	requested_attr.fileattr = ATTR_FILE_DATALENGTH;
 
 	*fallback = false;
 
-- 
Mathematics is the supreme nostalgia of our time.