2013/7/18 Mateusz Loskot
on Wed Jul 17 2013, Mateusz Loskot
wrote: On 17 July 2013 04:48, Rene Rivera
wrote: On Sat, Jul 6, 2013 at 8:24 AM, Dave Abrahams
wrote: on Fri Jul 05 2013, Rene Rivera
wrote: I started looking at making the final minor set of changes to the regression testing scripts to have the option of testing the modular git
version of the boost super-project. Unfortunately it now seems that
super-project no longer has *any* of the references to the library sub-projects. Me not being able to get this work done now is rather annoying as it likely means that the testing will not be ready early in the 1.55 release cycle. Which is the release currently scheduled to switch to git.
What is the status of the git transition?
Sorry for the inconvenience. The garbled results we were getting a little while back led us to analyze what the tool we were using was doing, and we realized the logic we inherited in that codebase was just wrong. So we've been rewriting the guts of that tool mostly from the ground up (https://github.com/ryppl/Boost2Git/pulse).[1] Now the contents of the inidividual repositories look basically sane.
When will the Boost release team be able to start working on the switch?
Porting the submodule logic that I had already added to the inherited codebase is the touch I hope to put on it this week (maybe even over
On 18 July 2013 08:22, Dave Abrahams
wrote: the the weekend).
And, believe me, I want to be done with this as much or more than anyone, so I'm going as fast as I can. If anyone would care to pitch in, we'd love to have you.
OK, got a chance to try a full checkout.. And ran into an infinite recursion in dulwich while getting the geometry repo. It looks like something is messed up with the mappings for that one. When I go look at it on github directly I see duplicate "include" directories at the root < https://github.com/boostorg/geometry>.
I just managed to clone it with GitHub for Windows. The problem is that the include directory with last commit 2 years ago is prefixed with whitespace, so it is " include", whereas the latter is "include" (screenshot attached, not sure if it will get through).
I guess that GitHub trims the blank characters before rendering HTML, so it is not visible.
and, I guess, Dulwich can't cope with the space(?)
We can easily map " include" in SVN into "include" in Git if that's the appropriate thing to do here.
Possibly, I don't understand where this whitespace come from, but would it be better to simply remove the " include" directory?
The whitespace came from an error in the ruleset. Fixed here: https://github.com/ryppl/Boost2Git/commit/1ff4eb06a986f6594d2b14504a0adb90cb... cheers, Daniel