@Aurélien, I edited your report so that the code showed rather than being interpreted.

@Eric, this fix seems simple enough, though I'm not sure why the postgres adapter needs an explicitely cast string. The other thing bothering me is it seems wikified text doesn't get sanitized correctly, stuff in "< % % >" seems to either get thrown out or (worse) rendered.
--------------------------------------------------------------------------------
@Felix
Thank you, I wasn't aware of this formatting problem.

I grepped 'find_by_revision' in source and this comes out:

<pre>
app/views/repositories/_dir_list_content.rhtml:19:<% changeset = @project.repository.changesets.find_by_revision(entry.lastrev.identifier.to_s) if entry.lastrev && entry.lastrev.identifier %>
app/helpers/application_helper.rb:513: if project && (changeset = project.changesets.find_by_revision(identifier))
app/models/repository/darcs.rb:32: patch = identifier.nil? ? nil : changesets.find_by_revision(identifier)
app/models/repository/darcs.rb:37: patch = identifier.nil? ? nil : changesets.find_by_revision(identifier)
app/models/repository/darcs.rb:55: patch = identifier.nil? ? nil : changesets.find_by_revision(identifier.to_s)
app/models/repository/darcs.rb:60: patch_from = changesets.find_by_revision(rev)
app/models/repository/darcs.rb:62: patch_to = changesets.find_by_revision(rev_to) if rev_to
app/models/repository/cvs.rb:33: rev = identifier.nil? ? nil : changesets.find_by_revision(identifier)
app/models/repository/cvs.rb:38: rev = identifier.nil? ? nil : changesets.find_by_revision(identifier)
app/models/repository/cvs.rb:43: change=changes.find_by_revision_and_path( entry.lastrev.revision, scm.with_leading_slash(entry.path) )
app/models/repository/cvs.rb:57: rev = identifier.nil? ? nil : changesets.find_by_revision(identifier)
app/models/repository/cvs.rb:64: changeset_from=changesets.find_by_revision(rev)
app/models/repository/cvs.rb:66: changeset_to=changesets.find_by_revision(rev_to)
</pre>

Maybe those also need to be escaped for postgresql ? I'm sure since I do not use darcs neither cvs, I can't say if it is a problem or not.
--------------------------------------------------------------------------------
Seems like Postgres 8.3 and 8.4 differ in the way the interpret data types. Especially by interpreting strings as integers and vice versa. We have seen various such issues in the last few days.
--------------------------------------------------------------------------------
Anyone have access to a Postgres 8.3 database to test this out on? I'm on 8.4 and don't want to fight a downgrade if I can help it.
--------------------------------------------------------------------------------
I can, I'll give you a feedback in the day.
--------------------------------------------------------------------------------
I did not have the time to test it deeply, but it seems it works fine with Postgresql 8.3. Maybe I'll have a bit more time tomorrow evening (french hours, it's 19h30 here!) to explore it if you want.
--------------------------------------------------------------------------------
Postgres 8.3 works as advertised with and without the original patch.
--------------------------------------------------------------------------------
I'm closing this as can't reproduce.

Aurélien Pocheville, once Redmine 1.0 is released could you upgrade and let us know if this is still happening for you? If so, feel free to reopen the issue.
--------------------------------------------------------------------------------

relates,Closed,3942,Redmine + PostgreSQL 8.3: ActiveRecord::StatementInvalid in RepositoriesController#revision