[PATCH 2 of 2] pager: if old pager extensions is enabled, respect pager.attend

[Martin von Zweigbergk]

On Mar 14, 2017 2:05 AM, "Yuya Nishihara" <yuya at tcha.org> wrote:

On Tue, 14 Mar 2017 00:03:53 -0700, Martin von Zweigbergk via
Mercurial-devel wrote:
> # HG changeset patch
> # User Martin von Zweigbergk <martinvonz at google.com>
> # Date 1489466597 25200
> #      Mon Mar 13 21:43:17 2017 -0700
> # Node ID 4b4b0fb20fc768c1ea6b330a5f37f3e325e88b32
> # Parent  d17402c07698443c13b082df8a8635a82b83d097
> pager: if old pager extensions is enabled, respect pager.attend
>
> This patch makes us respect pager.attend again if the extension is
> enabled. It also brings back the default attend list, so e.g. summary
> is not paged by default, just like it used to be before pager was
> moved into core.

Does it mean we'll support the legacy pager extension forever? That might
cause
confusion as new users will be likely to see new pager behavior, but old
users
would be too lazy to update their hgrc.


We have traditionally been very conservative about BC and I'm worried we're
breaking it too much without this patch.

We had a report internally about this breakage when we released a new hg
version yesterday. We have a log alias called xl that we also have enabled
paging for by default. Our user was setting pager.attend=<all our defaults,
except xl> as a way of not getting xl paged. When we upgraded to the new
hg, without disabling the extension, his "hg xl" output started getting
paged. It also wasn't obvious how to get back the old behavior. He tried
setting pager.attend-xl=no, but that only works for non-aliases. Setting it
in pager.ignore also doesn't work until they have disabled the pager
extension (and it requires it to be "log", not "xl"), and requiring two
simultaneous changes makes it really tricky. Kyle and I spent around 20
minutes trying to figure out the right way of getting his behavior back to
what he wanted. Also, it wasn't exactly the same as before, since all log
aliases would now be treated the same. That last thing is usually
desirable, so probably a good thing, but I do think the other pieces were
too complicated for users.

I'm not sure we need to support the legacy pager forever, but I do think we
need to make the transition smoother. I'm much more concerned about
changing behavior for existing users than the confusion about existing
users and new users getting different behavior. Even if we do end up
supporting the legacy pager forever, it's only about 50 lines of code, and
it's pretty isolated in pager.py.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.mercurial-scm.org/pipermail/mercurial-devel/attachments/20170314/fa457b09/attachment.html>