[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: SQLite and the Ruby bindings tests on Windows

From: Kouhei Sutou <kou_at_cozmixng.org>
Date: Wed, 10 Dec 2008 21:16:13 +0900 (JST)


In <ae6cb1100812082230u777205e8l60db5e0ee161732d_at_mail.gmail.com>
  "SQLite and the Ruby bindings tests on Windows" on Mon, 8 Dec 2008 22:30:10 -0800,
  Joe Swatosh <joe.swatosh_at_gmail.com> wrote:

> Hi,
> I'm trying to find the best way to deal with the Ruby bindings test failures
> I'm experiencing when running on Windows. I've seen comments on the list to
> the effect that 4 or so tests are failing for others, but almost every test is
> failing for me right now so I've been no help trying to trace anything down.
> The following script using the Ruby bindings runs without a problem in r33729
> on my WindowsXP system:
> require 'svn/repos'
> require 'svn/client'
> require 'fileutils'
> FileUtils.rmtree 'demo' if File.exist? 'demo'
> FileUtils.mkdir 'demo'
> wc_path = 'demo/wc'
> repos_path = 'demo/repos'
> repos_uri = "file:///#{File.expand_path(repos_path)}"
> file_path = "#{wc_path}/x.x"
> Svn::Repos.create(repos_path).close
> ctx = Svn::Client::Context.new
> ctx.set_log_msg_func { |items| [true, 'log msg'] }
> ctx.add_username_prompt_provider(0) {}
> ctx.auth_baton[Svn::Core::AUTH_PARAM_DEFAULT_USERNAME] = 'demo'
> ctx.checkout(repos_uri, wc_path)
> File.open(file_path, "w"){|f| f.print('content')}
> ctx.add(file_path)
> ctx.commit(wc_path)
> Svn::Repos.delete(repos_path)
> But fails in r33743 (the first revision that would build for me after r33730)
> with this message:
> D:\SVN\src-trunk\subversion\bindings\swig\ruby>demo.rb
> ./svn/util.rb:86:in `svn_repos_delete': (Svn::Error::SvnError)
> D:\SVN\src-trunk\subversion\libsvn_subr\io.c:1676: 720032: Can't
> remove file 'demo\repos\db\rep-cache.db': The process cannot access
> the file because it is being used by another process. from
> ./svn/util.rb:86:in `delete'
> from D:/SVN/src-trunk/subversion/bindings/swig/ruby/demo.rb:25
> And a more current trunk (r34604) just moves this lines in the back trace:
> D:\SVN\src-trunk\subversion\bindings\swig\ruby>demo.rb
> ./svn/util.rb:86:in `svn_repos_delete': (Svn::Error::SvnError)
> D:\SVN\src-trunk\subversion\libsvn_subr\io.c:1740: 720032: Can't
> remove file 'demo\repos\db\rep-cache.db': The process cannot access
> the file because it is being used by another process. from
> ./svn/util.rb:86:in `delete'
> from D:/SVN/src-trunk/subversion/bindings/swig/ruby/demo.rb:25
> I think this is because pool clean up that closes the cache doesn't run until
> after the end of the script and windows doesn't allow deleting open files.
> I can think of three solutions that basically boil down to: change SVN, change
> the bindings, or change the bindings tests. To me they roughly correspond to:
> 1) The script above is correct and should be supported. In this case I think
> it likely that the svn client code would have to change. Perhaps something
> like the patch I attached in
> http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=94719
> would be a start.

> 2) The script above is mostly correct and should be supported if the context
> was closed in some way. In this case the bindings would have to change to
> pass a different pool from the context object to the add and commit methods,
> and a new method to close the context would be added which would destroy the
> related pools. If this is the best solution, I hope that kou can help out
> because the pool stuff is pretty opaque to me.

What about adding Svn::Client::Context#close and
Svn::Client::Context#initialize accepts a block?

pseudo code:
  class Svn::Client::Context
    def initialize(...)
      return unless block_given?

    def close

But I'm not sure that 'close' is a good name for that...

> 3) The script above is ridiculous because sane person would add or commit
> (list might be another story, but don't let that ruin the sentiment) to a
> repository they were about to delete (except maybe someone trying to test the
> bindings). In this case I'm not sure what to do. The above script still
> fails if the Svn::Repos.delete() is in an at_exit{} block. So maybe we'd have
> to run each test in its own process? Ouch.
> 4) Give up on trying to support the Ruby bindings for a windows build.


Received on 2008-12-10 13:16:47 CET

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.