(from my email to p5p)

Notes:

• The "mesmoke.pl" is the live copy running on my Win32 VM. The "mesmoke" copy is an untested attempt to remove the --split-smoke failure.

• The "prep_git() / split-smoke" stuff was a failed experiment back when I was having Git run on my Linux host and the smoke run on the Win32 VM.

• You'll need Cygwin git. Msysgit causes test failures.

• To run as a "smoke-me" client, the variables %COMMITTER_EMAIL% and %COMMITTER_NAME% are search/replaced in the Test::Smoke config hash
  from a config file with ".template" appended. (ex: smoke-me_smoke.config.template -> smoke-me_smoke.config)

• I've attached my configurations as examples. From what I remember,
  forest mode is required and there's a variable that needs added to the Test::Smoke $conf configuration hash... maybe 'adir'? I forget. Also, it needs configured to use "rsync" but have the rsync command be a no-op.

• There's no configuration file yet so many variables in the script need
  changed per deployment. Something to fix.

• On the second and subsequent runs it complains (in Win32 VM copy): Filehandle STDIN reopened as $dotpatch only for output at mesmoke.pl line 126. Filehandle STDIN reopened as $logfh only for output at mesmoke.pl line 177. I've yet to find the cause (Perl test suite closing filehandles?) but it doesn't seem to have a negative impact yet.

• To run for blead I do: perl.exe mesmoke.pl --pattern ^remotes/origin/blead$ --what blead for smoke-me it is simply: perl.exe mesmoke.pl

• You'll probably want to do: set CYGWIN=nodosfilewarning to silence a warning.