bug: hg pull fails on a bitbucket repo

bug: hg pull fails on a bitbucket repo

[Nick Owens]

our hg (which i think is about version 1.4) will fail to pull some
bitbucket hg repos.

cpu% ramfs
cpu% cd /tmp
cpu% hg clone -r tip https://bitbucket.org/oridb/mc
destination directory: mc
requesting all changes
adding changesets
adding manifests
adding file changes
added 4988 changesets with 12508 changes to 1680 files
updating working directory
1090 files updated, 0 files merged, 0 files removed, 0 files unresolved
cpu% cd mc
cpu% hg pull
pulling from https://bitbucket.org/oridb/mc
searching for changes
abort: '/mnt/web/0/body' 400 Bad Request: /mnt/web/0/body
cpu%

hg is issuing a particular request where the resulting URI is over
4096 bytes, and the bitbucket server rejects this with a 400, and the
response body claims that this is because a line in the request is too
long (the URI). it's about 5200 bytes long in this instance.

in ports, there is mercurial 3.4.1
(https://code.9front.org/hg/ports/file/d0100ff1c37e/dev-vcs/mercurial/mkfile)
which does not have this problem. in newer hg, the wire protocol seems
quite different, but there's also a particular feature where hg can
send rpc arguments through split-up http headers rather than all
smushed into a URI.

there's a few options:

update the system hg to mercurial 3.4.1 (or whatever the latest
working copy is) and deal with the resulting fallout.
cherry-pick code from newer mercurial.
ignore the problem.

one caveat of updating to newer mercurial is that upstream dropped
python 2.5 support some years ago.

discuss.

Re: [9front] bug: hg pull fails on a bitbucket repo

[hiro]

> one caveat of updating to newer mercurial is that upstream dropped
> python 2.5 support some years ago.

so then how does your port work then?!

Re: [9front] bug: hg pull fails on a bitbucket repo

[Nick Owens]

you'd want 3.4.1 if you want newer mercurial features, or want to
avoid this bug.

mercurial 3.4.1 was release before they axed python 2.5 support.

Re: [9front] bug: hg pull fails on a bitbucket repo

[hiro]

fucking hell, so 3.4.1 is also already multiple years old?!

i doubt anybody would like to waste a week updating both python and hg.
also: how often will this game need to be repeated in the future if we
start with this now?

Re: [9front] bug: hg pull fails on a bitbucket repo

[Nick Owens]

our hg is approximately 9 years old. 3.4.1 is about 4 years old.

perhaps within another 5 years, more stuff will break. i and a
nontrivial amount of other people use bitbucket for mercurial hosting
and interact with it from 9front, so having it work reliably is pretty
desirable.
On Wed, Nov 28, 2018 at 3:44 PM hiro <23hiro@gmail.com> wrote:
>
> fucking hell, so 3.4.1 is also already multiple years old?!
>
> i doubt anybody would like to waste a week updating both python and hg.
> also: how often will this game need to be repeated in the future if we
> start with this now?

Re: [9front] bug: hg pull fails on a bitbucket repo

[hiro]

let's just wait 9 more years, perhaps bitbucket will stop supporting hg then.

Re: [9front] bug: hg pull fails on a bitbucket repo

[Kurt H Maier]

On Thu, Nov 29, 2018 at 02:02:03AM +0100, hiro wrote:
> let's just wait 9 more years, perhaps bitbucket will stop supporting hg then.

there are three editions of bitbucket, and only one supports hg any
more. I'd say 9 years from now we'll definitely be on our own.  

khm

Re: [9front] bug: hg pull fails on a bitbucket repo

[hiro]

i want my money back, mischief. nontrivial myass