I wanted to get rid of "from the developers" here, because
the PPA isn't package by Git developers. It's done by
third-party Ubuntu folks. I have no reason to think they
won't do a good job of keeping up to date, but I think it
makes sense to keep it clear that it's a third party.
This files are normally generated by the Makefile for
documentation. For building the manpages for the site, the logic is
quite similar but implemented in ruby.
* the tag_list is the full recursive tree. It is filtered afterwards for
some interesting files.
* the doc_list is the filtered list of main subject files.
The index task is in charge of gathering the documentation from the
git project and injecting it into database. Until now there was two
distinct files for retrieving the git project, one, `index.rake` which
allowed to get the files'content from github, and another one,
`local_index.rake` which allowed to get the content from a git
repository on the local disk. Both share the same logic in building
the pages and only differ on how to access them.
The purpose of this commit is too reunit the common logic. From now
on, only one logic of page building is provided.
This matches the change done to the preindex task in
5f7ca46e95. This shouldn't
affect most documents we build, since asciidoctor assumes
them to be manpages based on their content. But it does shut
up a bunch of "section 0" warnings when building
user-manual.txt.
I can't detect any visual change to the results, so this is
at least (hopefully) not regressing anything.
The "\1" and "\2" here are regex replacements, but the
double-quotes mean they get interpolated to the literal
bytes "1" and "2". We could fix this with "\\1", but it's
simpler still to just use a single-quoted string.
We can also get rid of the backslash-escaping of the
brackets. They're not needed on the replacement side of the
regex (and in fact we must get rid of them to not have them
show in the non-interpolated string).
This matches the recent change to preindex; otherwise our
fixed-up links go nowhere.
Note that in addition to tweaking the regex, we have to do a
few things to handle the subdirectory:
- we need to tell `ls-tree` to recurse now (the preindex
reads from octokit, which recurses by default)
- we need to shrink the path to its basename to match
preindex's flat view
The technical subdirectory is added to the list of processed
files. But the relative paths can not be processed correctly because
the route matching in the webserver doesn't expect subdirectories. So
we set everything at the same level.
The original seounded as if it wanted to teach how to subscribe to
the public-inbox archive, which does not make much sense. Separate
the two and then explain the benefit of subscribing a bit better by
taking suggestion made by Ævar on the mailing list (<878tkjk7m7.fsf@gmail.com>).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
I'm not sure which order the entries come in by default, but
using Versions.all.last seems to give 2.8.1, which is weird.
Using latest_version is clearly what we're going for.
I'm not sure why the $SEARCH_INDEX_URL variable is used. It
does provide a layer of indirection, and the existing site
has both Bonsai and Searchify set up. So maybe it was for
switching between them.
These days we're just using Bonsai, so let's point it at the
variable that is automatically set up by the Heroku add-on.
Jeff King
Felix Zeeb
dragon788
Jean-Noel Avila
Andrew Hayward
Pedro Rijo
Fabrice Salvaire
Alexandre Garnier
bedhanger
joernchen
Ben Abbott
Samuel Lijin
Jonathan Arbely
Frantret
Ævar Arnfjörð Bjarmason
Junio C Hamano
janusz