Yuya Nishihara <yuya@tcha.org> [Wed, 11 Feb 2015 19:57:07 +0900] rev 24108
shellquote: fix missing quotes for empty string
"hg kdiff3 -rREV" did not work because 1642eb429536 and 5edb387158a1 failed
to handle empty argument.
Martin von Zweigbergk <martinvonz@google.com> [Fri, 06 Feb 2015 16:09:43 -0800] rev 24107
trydiff: transpose 'if opts.git or losedatafn' with 'if f[12]'
Perhaps it's more readable this way...
Martin von Zweigbergk <martinvonz@google.com> [Thu, 22 Jan 2015 23:29:00 -0800] rev 24106
trydiff: extract function that generates filename pairs
The code that identifies copies/renames, as well as the filenames
before and after, is now isolated and we can extract it to a function
so it can be overridden by extensions (in particular the narrow clone
extension).
Martin von Zweigbergk <martinvonz@google.com> [Thu, 22 Jan 2015 23:18:43 -0800] rev 24105
trydiff: read file data in only one place
This moves getfilectx() out of the initial block in the loop, leaving
that block to be only about finding pairs of filenames in ctx1 and
ctx2 to diff.
Martin von Zweigbergk <martinvonz@google.com> [Thu, 22 Jan 2015 23:13:48 -0800] rev 24104
trydiff: set filename variables to None even when not opts.git or losedatafn
There is not much left of the first block "if opts.git or losedatafn"
block now. The next patch will move the call to getfilectx() out of
that block. We will then be using the defined-ness of 'f1' to tell
whether the file existed in ctx1 (and under what name). We will need
this information whether or not opts.git or losedatafn was set, so
just remove that guard. The only operation in the block that is not
cheap is the call to getfilectx(), but that has an extra 'if opts.git'
guard already.
--ignore-space-change proves that only 'if opts.git or losedatafn:'
was removed.
Martin von Zweigbergk <martinvonz@google.com> [Thu, 22 Jan 2015 23:01:00 -0800] rev 24103
trydiff: read flags in one place
Now that we can trust f1/f2 to indicate whether that side of the diff
exists, we can move the calls to ctx.flag() to a single place.
Martin von Zweigbergk <martinvonz@google.com> [Fri, 06 Feb 2015 15:21:56 -0800] rev 24102
trydiff: break 'if opts.git or losedatafn' into two
This moves the initialization 'binary' closer to its other assignment,
but exists mostly to simplify future patches.
Martin von Zweigbergk <martinvonz@google.com> [Thu, 22 Jan 2015 22:42:35 -0800] rev 24101
trydiff: make filenames None when they don't exist
f1 and f2 are currently set always set to some filename, even for
added or deleted files. Let's instead set them to None to indicate
that one side of the diff doesn't exist. This lets us use the filename
variables instead of the content variables and simplify a bit since
the empty string is not a valid filename. More importantly, it paves
the way for further simplifications.
Augie Fackler <augie@google.com> [Tue, 10 Feb 2015 10:57:58 -0500] rev 24100
test-help: add test to demonstrate that 'hg help merge-tools' is sane
Gregory Szorc <gregory.szorc@gmail.com> [Mon, 09 Feb 2015 23:07:39 -0800] rev 24099
help.merge-tools: do not double document merge tools
Merge tools were being double documented in help system output due
to functions being defined under multiple names in the merge tools
dictionary.
Establish a new dictionary for just the tools to document and
use it from the help system so we don't get double output.
Double documentation likely plagues other auto-documented items
as well. It might be a good idea to eventually compare function
instances to filter out duplicate entries from dictionaries
passed to ``makeitemsdoc``. However, without an easy way to break
ties, this may result in some functions being advertised over
their modern equivalents. This would be a noble patch series.
But it isn't one this author is willing to tackle at this time.
Gregory Szorc <gregory.szorc@gmail.com> [Mon, 09 Feb 2015 14:59:04 -0800] rev 24098
help: teach topic symbols how to dedent
When using docstrings for documenting symbols such as revsets,
templates, or hgweb commands, documentation likely has leading
whitespace corresponding to the indentation from the Python source
file.
Up until this point, the help system stripped all leading and
trailing whitespace and replaced it with 2 spaces of leading
whitespace. There were a few bad side-effects. First, sections
could not be used in docstrings because they would be indented
and the rst parser would fail to parse them as sections. Also,
any rst elements that required indentation would lose their
indentation, again causing them to be parsed and rendered
incorrectly.
In this patch, we teach the topic symbols system how to dedent
text properly. I argue this mode should be enabled by default.
However, I stopped short of changing that because it would cause
a lot of documentation reformatting to occur. I'm not sure if
people are relying on or wanting indentation. So, dedenting has
only been turned on for hgweb symbols. This decision should be
scrutinized.
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 06 Feb 2015 22:25:40 -0800] rev 24097
webcommands: document "graph" web command
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 06 Feb 2015 22:19:59 -0800] rev 24096
webcommands: document "archive" web command
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 06 Feb 2015 22:11:54 -0800] rev 24095
webcommands: document "filelog" web command
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 06 Feb 2015 22:08:30 -0800] rev 24094
webcommands: document "annotate" web command