More follow-up....
I've tracked it down further. HistoryCache.get() calls the
SubversionHistoryParser.parse, which checks repository.isIgnored()
When OpenGrok initialises and sets up this repo, it does:
if (!RuntimeEnvironment.getInstance().isRemoteScmSupported()) {
// try to figure out if I should ignore this repository
try {
SVNClient client = new SVNClient();
...
I have remoteScmSupported in my config.xml set to true, but for some
reason OpenGrok is picking it up as false and is going down and setting
this repository to be ignored.
cheers,
steve
Stephen Lau wrote:
> Stephen Lau wrote:
>
>> J?rgen Austvik wrote:
>>
>>
>>> Stephen Lau wrote:
>>>
>>>
>>>> I just deployed OpenGrok built from trunk and so far most things seem
>>>> to be working well, with one notable exception: Being able to view
>>>> the history of directories (both root and subdirs) within the SVN repo.
>>>> Anyone have any ideas?
>>>>
>>>>
>>> The 18th Trond said this in the "History and Annotate files in
>>> Subversion" thread:
>>>
>>> ---------8<-----------8<-----------8<-----------8<-----------8<-----------8<--
>>>
>>>
>>> The SubversionHistoryParser is buggy when it comes to parsing history
>>> for directories, so it will not populate the appropriate history
>>> entries to the various files (see
>>> http://src.opensolaris.org/source/xref/opengrok/trunk/src/org/opensolaris/opengrok/history/Repository.java#115
>>>
>>> )
>>> ---------8<-----------8<-----------8<-----------8<-----------8<-----------8<--
>>>
>>>
>>>
>>> Is it a symptom of the same problem?
>>>
>>> -J
>>>
>>>
>> My understanding was that this would prevent OpenGrok from being able to
>> cache the Subversion history for a directory, but that it would fallback
>> to pulling the history live (which is what it did previously, when I
>> last deployed OpenGrok from a trunk build on 2008-04-08).
>>
>>
> Best I can tell is that in history.jsp:58, it attempts to do
> HistoryGuru.getHistoryReader(f) where f is my directory path.
> This in turn calls getDirectoryHistoryReader, which tries to get a
> directoryHistoryParser which succeeds. Where it fails is
> HistoryGuru.java:156 which attempts to do historyCache.get(path)
>
> Since it can't build a history cache for Subversion directories, this
> fails - but it's a silent failure since there's no else clause for the
> "if (history != null)" immediately following, which ends up in an Error
> 404 (enoent) sent back to the webapp.
>
> Does that make sense?
> I would think that the old behaviour should still be applicable here:
> try the history cache, and if there is no entry in the history cache for
> the given path - fall through and do a real-time query to build the history.
>
> cheers,
> steve
>
>
--
stephen lau | stevel at opensolaris.org | www.whacked.net
