Re: [PATCH v3 06/10] clone: delay cloning until after remote HEAD checking

From: Nguyen Thai Ngoc Duy <pclouds@gmail.com>
Date: Jan 24, 2012, 07:01 UTC
Message-ID: <CACsJy8C_Lzgk-W3D9ArTh4xQBgFcy41UQzTHfghYKrGg7WxHMg@mail.gmail.com>
In-Reply-To: <7vvco253xx.fsf@alter.siamese.dyndns.org>

2012/1/24 Junio C Hamano <gitster@pobox.com>:

Show 14 quoted lines

> Recently, 6f48d39 (clone: delay cloning until after remote HEAD checking,
> 2012-01-16) tried to record if a remote helper needs to be called after
> parsing the remote when transport_get() is called, by overwriting the
> field meant to store the configured remote helper name in the remote
> structure.
>
> This is OK when a remote represents a single remote repository, but fails
> miserably when pushing to locations with multiple URLs, like this:
>
>    $ cat .git/config
>    [remote "origin"]
>        url = https://code.google.com/p/git-htmldocs/
>        url = github.com:gitster/git-htmldocs.git
>        push = refs/heads/master:refs/heads/master

My bad. remote.*.vcs (or remote->foreign_vcs) is designed to override remote helper detection based on url. Once you have pushed over https, the following urls will be over https too. Luckily no other places perform an operation over multiple urls like push so we're safe with your bandage.

We should have another way to detect whether remote helper being used (comparing some function pointers like transport->connect or transport->disconnect may help). But I need to think a bit about delaying cloning for http(s) too even though remote helper is used.

-- 
Duy