Alexander Litvinov [off-list ref] writes:
I have tried your test and found that cloning such repo does not clone
subprojects. Only empty dir 'sub' is being cloned.
Yes, that is pretty much intentional. Having a link for higher
layer tools can use (i.e. tree objects can contain 160000 mode
"gitlinks" to subproject commit objects) while not considering
such linkage part of the reachability is the whole point of
plumbing level subproject support. It allows people not to
download repositories of uninteresting subprojects. Higher
layer tools such as clone/checkout/diff could be instructed
(currently they cannot be, though) to recurse into subproject
directories if the user wants to.
A few good pointers for this topic are:
$gmane/44162
$gmane/44308
On 5/11/07, Junio C Hamano [off-list ref] wrote:
Alexander Litvinov [off-list ref] writes:
quoted
I have tried your test and found that cloning such repo does not clone
subprojects. Only empty dir 'sub' is being cloned.
Yes, that is pretty much intentional. Having a link for higher
layer tools can use (i.e. tree objects can contain 160000 mode
"gitlinks" to subproject commit objects) while not considering
such linkage part of the reachability is the whole point of
plumbing level subproject support. It allows people not to
download repositories of uninteresting subprojects. Higher
layer tools such as clone/checkout/diff could be instructed
(currently they cannot be, though) to recurse into subproject
directories if the user wants to.
fwiw: I just released cgit 0.3, which includes a script
(submodules.sh) used to initialize, update and check status of
submodules. It uses the file .gitmodules to map between submodule path
and git repository url (cgit now uses git as a submodule).
You can clone the repo here:
git://hjemli.net/pub/git/cgit
just look at the files here:
http://hjemli.net/git/cgit/
--
larsh