Avoid network timeout on large histories

[brlcad]

This change should avoid a reverse svn log failure connection timeout by pulling the log in order (and extracting the last entry). This is only done if the reverse log fails in order to keep overhead low for most enlistments, but this should fix failures for a number of repositories that have a large revision history.

[PDegenPortnoy]

I don't see that run(back_to_front) will return a truthy value. Please refer to Scm::Adapters::AbstractAdapter. This calls Shellout#run, which calls Shellout.execute, which returns an array of [status, outbuf.string, errbuf.string]

It would be very helpful to have test coverage of this next_revision_xml method with the different pathways covered.

[brlcad]

The perils of writing code one cannot easily test. Fully concur that there's a problem as written. Looks like I misread a similar snippet in lib/scm/adapters/git/head.rb that let me to think a return code would result.

What about pushing the conditional into the command begin run, then making the callee always just check parse().last?
run("#{back_to_front} || #{front_to_back}")

I can redo a push with that change if the explanation is unclear.

[amujumdar]

@brlcad @peterdp since this repository is on SourceForge, I am going to try SvnSync'ing it. That might fix the issue altogether. Currently, we seem to be treating it as an SVN repository, one that cannot be SvnSync'ed and trying to convert it into a Git repository locally.

[josie559]

G