Danek Duvall <danek.duvall at sun.com> writes:
> On Sat, Aug 09, 2008 at 03:47:36PM -0400, Richard Lowe wrote:
>
>> The changes therein are:
>> - Shows all of "description", not just the first line
>> - Shows files one-per-line
>> - Uses 'modified', 'added', 'deleted', 'copied', rather than file
>> file+ file- etc, adds the word "from" in the parens in "copied"
>> entries.
>> - Adds spurious newlines in some cases, as per my initial mail.
>
> They're not spurious. The problem is this bit:
>
> {files}\n{file_adds}\n{file_dels}\n{file_copies}
>
> which adds three newlines (in addition to the two before the block and
> three after), regardless of whether there are any files modified, added,
> removed, or copied. The varying number of blank lines you're seeing is
> because some changesets have added or removed files and others don't. You
> want to have
>
> {files}{file_adds}{file_dels}{file_copies}
>
> and set
>
> end_file_adds = '\n'
>
> and so on. I have style quibbles with it (too much break between
> changesets, don't like the indented comments), but other than that, it
> works just fine.
>
> I'd be careful about using file_adds and file_dels, though, since the
> mercurial folks constantly warn about how slow it is to retrieve that data.
> On a single changeset, it's not really noticeable, but on the 450+
> changesets in the IPS gate, it goes from .15s to 7.8s to spit out the
> entire log.
Right, with onnv-gate it's the difference between 1.8s and 15 minutes
(and counting), that clearly won't fly.
Do we want just the {files} list in RTIs, or add/del/modify?
If the latter, are we happy telling people to use --debug -v?
-- Rich
