[Gnu-arch-users] transient archives?

[Jeremy Shaw]

Hello,

Is there a sane way to make temporary archives?

I am imagining doing something like this:

I want to work on a branch of
'address@hidden/project--mainline--1.0'. So I do the
following steps:

(1) Use a tool like pbuilder to build a clean chroot-able development
    environment that contains only the needed build-dependencies +
    editor + tla

(2) create a temporary archive in the build environment and tag from
    'address@hidden/project--mainline--1.0' to
    'address@hidden/project--devo--1.0'

(3) Hack things up

(4) Submit relevant changesets from chroot archive to
    address@hidden/project--mainline--1.0 via PQM.

(5) Wait for confirmation of successful integration into mainline.

(6) Delete chroot environment

Clearly an important issue is the uniqueness of the revision names.

If I do that same procedure twice, using the same name for the chroot
archive, then I will create two sets of revisions with identical
names, but different changesets. That is clearly bad.

One solution is to ensure that the temporary archive name is uniquely
generated each time. For example:

tla make-archive  address@hidden( uuidgen )

If I had write access to the mainline, I could also doing something
along these lines:

(1)-(3) the same as before

(4) tla sync-tree project--devo--1.0--base-0 (the remove the patchlogs
    that refer to the temporary archive -- note that this does not
    actually work, since sync-tree only adds patchlogs)

(5) tla changes -o ,,feature-xyz

(6) tla move ,,feature-xyz ../project--mainline--1.0 && cd 
../project--mainline--1.0

(7) tla redo ,,feature-xyz

(8) $EDITOR $( tla make-log  )

(9) tla commit

Now I don't have any patch-logs from the temporary archive. Also, all
my hacking is neatly packaged in a single patch in mainline (which may
or may not be a good thing). Of course, if anyone (myself included)
has tagged off of project--devo--1.0, those branches are in a bit of
limbo.

Of course, you may question why I want a temporary archive at all. I
am trying to design a system, where people can benefit from the
decentralization features of tla, but I don't want to burden them with
having to maintain archives on their local machines in a reliable
fashion.

For example, if someone goes on "vacation" (with out internet access),
and takes their laptop, I want them to be able to make a local branch
on the laptop (bofore they leave), make more branches off that branch,
commit changes regularily -- and then when they come back, merge their
changes into a carefully maintained, backed up archive. They should
then be able to wipe the laptop clean with a fresh install.

Any thoughts?

Jeremy Shaw.