On 2016-11-17 08:57, Jef Driesen wrote:
Yes, there is the mirror at sourceforge: git clone git://git.code.sf.net/p/libdivecomputer/code
There still is something off with the git.libdivecomputer.org-repo. If I do:
$ git clone git://git.libdivecomputer.org/libdivecomputer.git Cloning into 'libdivecomputer'... remote: Counting objects: 7881, done. remote: Compressing objects: 100% (2326/2326), done. remote: Total 7881 (delta 6319), reused 6864 (delta 5551) Receiving objects: 100% (7881/7881), 1.50 MiB | 969.00 KiB/s, done. Resolving deltas: 100% (6319/6319), done. Checking connectivity... done. warning: remote HEAD refers to nonexistent ref, unable to checkout.
$ cd libdivecomputer $ git checkout master error: pathspec 'master' did not match any file(s) known to git.
$ git ls-remote origin 4df5e490083fe235293d3c17903b48d56100775a refs/heads/release-0.1 154fc82da1f249674cd1e3dd55b8aa0650eb7b2d refs/heads/release-0.2 1a3c919a8255cc35d05d3345a4e8531c38a583ca refs/heads/release-0.3 9f42f6b87d89b9957b266896e75aea3b16d86ab2 refs/heads/release-0.4 a5b009ed0c54b0b57e85735dea82db358b949dcc refs/tags/v0.1.0 c3ea5a16e49224a3007ee4f37b58bae594a84e1b refs/tags/v0.1.0^{} 653ec757a53d3056f24b04537d6bbd29f6de6741 refs/tags/v0.2.0 2ed54a204bb8f8167736505c02747d02a7a02fc9 refs/tags/v0.2.0^{} a52f856209d0b2aa022151471b1473c71bfb9c4c refs/tags/v0.3.0 82fadd89bfe16918904d47558dcea2bb1afcfbc1 refs/tags/v0.3.0^{} 26de43ef6bfba56fe13c3f1c36709ad3d710aed9 refs/tags/v0.4.0 bdd8e06786806bbc2dfa3b2628b7945b3ce25e21 refs/tags/v0.4.0^{} 4e03cd2d6237027a42baa44deadfc005d8542f19 refs/tags/v0.4.1 2d3e7078394f128d08b676e3d59e0f9f3e02963e refs/tags/v0.4.1^{} 00ac0105ffbd52ff30df027ad06ebb4e451c52c5 refs/tags/v0.4.2 9f42f6b87d89b9957b266896e75aea3b16d86ab2 refs/tags/v0.4.2^{} 67f05c38e2172b7b85bc2cfb05c1507e7d861773 refs/tags/v0.5.0 ba0e7224b0fd68fe319b0a0ffb4caaf721b3148b refs/tags/v0.5.0^{}
While the clone from Sourceforge says:
$ git ls-remote origin 296b13947f2f002929d60aaa29b1d686fc813d55 HEAD 296b13947f2f002929d60aaa29b1d686fc813d55 refs/heads/master 4df5e490083fe235293d3c17903b48d56100775a refs/heads/release-0.1 ...
Should the Sourceforge be considered to be the primary source?
Poltsi
On 2016-12-07 13:09, Paul-Erik Törrönen wrote:
On 2016-11-17 08:57, Jef Driesen wrote:
Yes, there is the mirror at sourceforge: git clone git://git.code.sf.net/p/libdivecomputer/code
There still is something off with the git.libdivecomputer.org-repo. If I do:
$ git clone git://git.libdivecomputer.org/libdivecomputer.git Cloning into 'libdivecomputer'... remote: Counting objects: 7881, done. remote: Compressing objects: 100% (2326/2326), done. remote: Total 7881 (delta 6319), reused 6864 (delta 5551) Receiving objects: 100% (7881/7881), 1.50 MiB | 969.00 KiB/s, done. Resolving deltas: 100% (6319/6319), done. Checking connectivity... done. warning: remote HEAD refers to nonexistent ref, unable to checkout.
$ cd libdivecomputer $ git checkout master error: pathspec 'master' did not match any file(s) known to git.
$ git ls-remote origin 4df5e490083fe235293d3c17903b48d56100775a refs/heads/release-0.1 ...
While the clone from Sourceforge says:
$ git ls-remote origin 296b13947f2f002929d60aaa29b1d686fc813d55 HEAD 296b13947f2f002929d60aaa29b1d686fc813d55 refs/heads/master 4df5e490083fe235293d3c17903b48d56100775a refs/heads/release-0.1 ...
I can confirm the problem. It looks like the HEAD and master refs are missing. The strange part is that I only see this problem with the git protocol, not with the ssh protocol. (And that's why I never noticed this myself.)
@Dirk: Can you check the server setup?
Should the Sourceforge be considered to be the primary source?
The git repository on libdivecomputer.org is the primary repository. But I keep the sourceforge repo in sync as a mirror. So unless there is something wrong with one of them, there shouldn't be any difference.
Jef
On Dec 7, 2016, at 05:05, Jef Driesen jef@libdivecomputer.org wrote:
I can confirm the problem. It looks like the HEAD and master refs are missing. The strange part is that I only see this problem with the git protocol, not with the ssh protocol. (And that's why I never noticed this myself.)
@Dirk: Can you check the server setup?
Yes, it's a permissions problem where for some strange reason pushes to the repo via ssh set the wrong permissions on the objects they create, including the HEAD ref - so that can no longer be read via non-ssh connections.
I thought I had that fixed (and it seems to work again for Subsurface), but clearly it's still hosed for libdivecomputer. I'm traveling which makes this a bit more annoying (stupid 2FA token is at home), but I'll try to fix this ASAP.
Should the Sourceforge be considered to be the primary source?
No, Dirk should be considered an incompetent server admin. But as Jef might say "the price is right" (i.e., I do all the work, I pay for the servers, for the data center cost, for the bandwidth, etc). At this time I am achieving something like 97% uptime with working services, I guess :-)
The git repository on libdivecomputer.org is the primary repository. But I keep the sourceforge repo in sync as a mirror. So unless there is something wrong with one of them, there shouldn't be any difference.
Same for Subsurface. git.subsurface-divelog.org is the primary repository, there are two github repos (Linus' and mine) that I also push to as backup.
/D
On 2016-12-07 15:05, Jef Driesen wrote:
I can confirm the problem. It looks like the HEAD and master refs are missing. The strange part is that I only see this problem with the git protocol, not with the ssh protocol. (And that's why I never noticed this myself.)
The git repository on libdivecomputer.org is the primary repository. But I keep the sourceforge repo in sync as a mirror. So unless there is something wrong with one of them, there shouldn't be any difference.
Ok, so would it be fair to say that currently (over the git-protocol) the Sourceforge can be considered to be the newest accessible version of the code?
I'm asking this because I'm attempting to rebase my VMS/Sentinel-stuff to the latest version of libdc. I think I made a mess of it as my code is not a branch but a clone of libdc-0.4, but the changes/additions are fairly simple.
Poltsi
On Dec 7, 2016, at 11:17 PM, Paul-Erik Törrönen poltsi@poltsi.fi wrote:
On 2016-12-07 15:05, Jef Driesen wrote:
I can confirm the problem. It looks like the HEAD and master refs are missing. The strange part is that I only see this problem with the git protocol, not with the ssh protocol. (And that's why I never noticed this myself.)
The git repository on libdivecomputer.org is the primary repository. But I keep the sourceforge repo in sync as a mirror. So unless there is something wrong with one of them, there shouldn't be any difference.
Ok, so would it be fair to say that currently (over the git-protocol) the Sourceforge can be considered to be the newest accessible version of the code?
This is fixed on git.libdivecomputer.org (and shouldn't happen again - famous last words)
/D
-
Dirk Hohndel -
Jef Driesen -
Paul-Erik Törrönen