changing $worksrcdir break the build when using git.fetch or hg.fetch

[ci42]

git/hg fetch the source code directly into $worksrcpath. If your port need to change the $worksrcdir (e.g. the source to be built is in a subdirectory) the build fails because $worksrcpath is defined as $workpath/$worksrcdir and the source is now in the wrong directory. See the attached portfile for an example that demostrates this behavior.

The suggested patch uses $workpath/$distname as the location for git/hg fetch.

[ci42]

use this to reproduce the bug

[ci42]

suggested patch

[ryandesign (Ryan Carsten Schmidt)]

Note that you don't need to change worksrcdir in your portfile. You could instead change configure.dir and/or build.dir, for example.

[ci42]

Replying to ryandesign@…:

Note that you don't need to change worksrcdir in your portfile. You could instead change configure.dir and/or build.dir, for example.

I know, and for this portfile this is already fixed (see #28799). I used this as an example because I encountered this bug while developing the portfile. The problem is there anyway.

[ryandesign (Ryan Carsten Schmidt)]

I don't think there's any need to wait until 2.0.0; I don't think changing this should break any existing ports. At the same time, we can fix fetch.type svn to have the same behavior; it's always bugged me that it behaved differently.

[raimue (Rainer Müller)]

I thought we do not want to do any 1.x release anymore? Other changes on trunk are currently targeting 2.0.

[ryandesign (Ryan Carsten Schmidt)]

Oh, in that case whatever. :)

[ryandesign (Ryan Carsten Schmidt)]

This needs to be fixed differently. hgfetch and gitfetch need to be rewritten to use command_exec (like bzrfetch, cvzfetch and svnfetch do) instead of system.

[ci42]

use command_exec for git.fetch

[ci42]

Replying to ryandesign@…:

This needs to be fixed differently. hgfetch and gitfetch need to be rewritten to use command_exec (like bzrfetch, cvzfetch and svnfetch do) instead of system.

Please have a look at the new patch. It uses the 'command_exec' for git.fetch and fetches the source to '$workpath/$distname' like the first patch. Feedback and suggestions are welcome.

[ci42]

Replying to ryandesign@…:

This needs to be fixed differently. hgfetch and gitfetch need to be rewritten to use command_exec (like bzrfetch, cvzfetch and svnfetch do) instead of system.

Please have a look at the latest patch. 'command_exec' is used for both git.fetch and hg.fetch. The source will also be fetched to '$workpath/$distname' like the first patch does. Feedback and suggestions are welcome.

[ci42]

use command_exec for git.fetch and hg.fetch

[tvb377@…]

Cc Me!

[cooljeanius (Eric Gallager)]

Cc Me!

[neverpanic (Clemens Lang)]

Let's check whether this is still a problem. If so, we'll consider it for 2.5.0 – maybe this is fixed with the VCS fetch caching changes.

[jmroot (Joshua Root)]

Ticket retargeted after milestone closed

[jmroot (Joshua Root)]

Ticket retargeted after milestone closed

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #45353.