Ramkumar Ramachandra wrote on Thu, Aug 19, 2010 at 02:38:25 +0530: > Daniel Shahaf wrote: > > svnrdump needs an inequality check here. But 'svnadmin load', for > > example, needs an "is at most" check. So I don't think this is generic > > enough: it would be better to provide something that 'svnadmin load' can > > use too. > > > > So, two options: > > > > * Repeat the parse_format_version() trick (from load.c) in svnrdump. > > It's actually not possible. If I read one line of the dumpstream in > load_editor.c and then call svn_repos_parse_dumpstream2, the function > won't have the version number information because I've already read > that out from the dumpstream, and I can't rewind the stream.
Yeah, I haven't slept properly in three days, and that's the result :(. When I made the suggestion, I thought parse_format_version() lives in svnadmin/main.c. Clearly I was hallucinating. > Note that as you already pointed out on IRC, parse_format_version > imposes a maximum already (load.c:647). > > > * In the API, PARSE_FNS could grow a dumpfile_version_record() callback > > (like it has a uuid_record()) callback. The caller can implement > > whatever check they want in that callback. (possibly with a special > > error code that svn_repos_parse_dumpstreamN() recognizes; compare > > usage of SVN_ERR_CANCELLED) > > Excellent idea; too much work for one patch though- I'll add a TODO > comment about this and re-post the patch. > I see you've posted a separate patch now. > -- Ram
