Hi,

Quoting Ian Jackson (2016-12-21 12:10:38)
> Sean Whitton writes ("Bug#848931: dgit: please add a configuration variable 
> that prevents me from accidentally doing a non-source-only upload"):
> > Good idea.  This option should be ignored when --new is present.
> 
> I was thinking about this some more and I wonder if a different
> approach would be better.
> 
> If we're doing a source-only upload, the user might not want to do
> binary-build-for-upload at all.  It is only binary-builds-for-upload
> that need to be done in a clean chroot: source packages can (I think,
> and dgit relies on this) safely be manipulated in the user's working
> environment.
> 
> So I was thinking that maybe there should be a `dgit push-source'
> subcommand which does not depend on the existence, or look for, any
> previously built artefacts.  It would build the source package afresh (a la
> dgit `build-source') and upload it, in one subcommand.

solving this issue like that might fix another issue I usually have.

I usually do:

 - dgit sbuild # checking that everything works fine
 - dgit build-source
 - dgit push

Unfortunately, between the first and the second step, I have to manually remove
the *.changes files that the "dgit sbuild" step left. I understand that this is
because there would otherwise be an ambiguity about which .changes to upload.

With a "dgit push-source" subcommand, this ambiguity would be gone and I might
no longer have to manually rm the .changes files created by sbuild.

Thanks!

cheers, josch

Attachment: signature.asc
Description: signature

Reply via email to