Bug#901310: gitaly hard codes gitlab username and it does not use the same options to adduser as gitlab package
On Mon, 11 Jun 2018 22:38:37 +1000 Dmitry Smirnov wrote: > On Monday, 11 June 2018 10:14:28 PM AEST Pirate Praveen wrote: > > I think the solution is to have a gitlab-common package that will handle > > the user creation with both gitlab and gitaly depending on it. > > Good idea. Please review https://salsa.debian.org/ruby-team/gitlab/commit/6eccbab22731b4fefd53161cf239d106e4ab6038 signature.asc Description: OpenPGP digital signature
Bug#901310: gitaly hard codes gitlab username and it does not use the same options to adduser as gitlab package
On Monday, 11 June 2018 10:14:28 PM AEST Pirate Praveen wrote: > I think the solution is to have a gitlab-common package that will handle > the user creation with both gitlab and gitaly depending on it. Good idea. -- Cheers, Dmitry Smirnov. signature.asc Description: This is a digitally signed message part.
Bug#901310: gitaly hard codes gitlab username and it does not use the same options to adduser as gitlab package
On June 11, 2018 3:25:35 PM GMT+05:30, Dmitry Smirnov wrote: >On Monday, 11 June 2018 6:52:59 PM AEST Pirate Praveen wrote: >> It should check the username from debconf db already set by gitlab >and >> also ask debconf question if not set already. > >It should not fail to install on the box without gitlab. It should >install >successfully even if gitlab was never installed. I think the solution is to have a gitlab-common package that will handle the user creation with both gitlab and gitaly depending on it. >IMHO this is not much value in debconf prompt for user name... It may >be nice >but perhaps would be excessive in most cases. > >I mean, "gitlab" user name is much like "www-data". Maybe it would be >best >not to alter it through debconf to avoid significant maintenance >overhead for >little value. It is significant because it forms the ssh URL and visible to every user of the service. g...@domain.tld is preferred by many whereas default is git...@domain.tld >How are you going to alter .service file? Are you sure it is worth the >trouble? Just how we did it for gitlab. > >> postrm should not remove the user if gitlab package is still present. > >I would only implement this so purge wouldn't remove "gitlab" user when > >gitlab is installed. gitlab-common as explained above can fix this. -- Sent from my Android device with K-9 Mail. Please excuse my brevity.
Bug#901310: gitaly hard codes gitlab username and it does not use the same options to adduser as gitlab package
On Monday, 11 June 2018 6:52:59 PM AEST Pirate Praveen wrote: > It should check the username from debconf db already set by gitlab and > also ask debconf question if not set already. It should not fail to install on the box without gitlab. It should install successfully even if gitlab was never installed. IMHO this is not much value in debconf prompt for user name... It may be nice but perhaps would be excessive in most cases. I mean, "gitlab" user name is much like "www-data". Maybe it would be best not to alter it through debconf to avoid significant maintenance overhead for little value. How are you going to alter .service file? Are you sure it is worth the trouble? > postrm should not remove the user if gitlab package is still present. I would only implement this so purge wouldn't remove "gitlab" user when gitlab is installed. -- Cheers, Dmitry Smirnov. --- Truth — Something somehow discreditable to someone. -- H. L. Mencken, 1949 signature.asc Description: This is a digitally signed message part.
Bug#901310: gitaly hard codes gitlab username and it does not use the same options to adduser as gitlab package
Package: gitaly version: 0.96.1+debian-3 severity: important It should check the username from debconf db already set by gitlab and also ask debconf question if not set already. postrm should not remove the user if gitlab package is still present. signature.asc Description: OpenPGP digital signature
