Sorry for the lag...
1. Why'd you specify branch & revision as transient? Not clearly wrong,
just seems unnecessary.
2. You're changing the branch output from my "just the last term" spec to
"full branch path from repo root," right? That's surely more correctly
complete, but may make the tag a bit unwieldy. It's got some ripple effects
in other tools we have, if we're changing spec. (I don't feel strongly,
though I'd originally take the last-term as "almost surely unique" and
mildly more usable.)
3. Making svnversion optional doesn't distress me (and I assume git can't
offer it), but it does potentially allow people to misrepresent builds as
being rNNN, when they're actually arbitrary local mods from NNN, perhaps
including split-version checkouts, but certainly including local edits. Do
we want the imprecision recorded somehow, if svnversion isn't present?
(This isn't entirely theoretical; both Ray and Joel have accidentally had
mixed-version workspaces, though only by branch-info.txt being off-base
w.r.t. the rest.)
On Sat, Mar 7, 2009 at 5:33 AM, Scott Blum <sco...@google.com> wrote:
> Hey Freeland,
> I hacked on SvnInfo a bit to do the following things:
>
> 1) If .svn/ isn't present but .git/ is, use "git svn info".
>
> 2) A more surefire way of figuring out exactly what branch we're on. Note
> the behavior change for a release branch: whereas before you'd get "1...@4444"
> you'll now get "releases/1...@4444". To me this seems like an improvement.
>
> 3) Make svnserve's success optional; use whatever rev you got from "svn
> info" if it doesn't run. I don't feel strongly about this bit, just thought
> it might be okay to be more lenient.
>
> Thanks,
> Scott
>
>
--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---
