[darcs-users] How to you handle content-related dependencies?

[Miles Gould]

On Mon, Jun 6, 2011 at 1:41 PM, Sittampalam, Ganesh
<ganesh.sittampalam at credit-suisse.com> wrote:
> You could also just run a minimal testsuite to infer what the
> dependencies are, instead of a procedure that's very tailored to the
> format of your content.

I don't follow - could you please be more explicit? And what sort of
testsuite would one write for a TeX project, anyway?

If you mean

1. Check out new repo
2. Pull just-committed patch into new repo
3. Pull more patches* until `thesis.dvi` builds without errors
4. Add dependencies on all the patches you had to pull

then it sounds rather slow**. Recording a new patch shouldn't be a "go
and make a fresh pot of coffee while you're waiting" exercise, IMHO.

* This is the interesting step, of course. If you wanted to be fully
general you could walk the patch-dependency graph to find a minimal
set of dependencies, but again this sounds rather slow; alternatively
you could scan the error messages for "missing foo" errors (in a
language-dependent way) and attempt to add dependencies accordingly.
Since (if I'm reading `darcs help changes` right) darcs has no
equivalent of `git log -S` (search for patches adding or removing a
given string) you'd still need to cache theorem-creation information
in the commit messages as I described in my last message.

** I dunno about you, but a clean build of my full thesis from source
took *ages*. Bloody xypic.

Miles.