Due to a bug in Maven Indexer (see URL), a repository index using the new GZIPped format cannot be downloaded by NB when the HTTP server sets Content-Encoding: gzip. http://jira.codehaus.org/browse/MINDEXER-12 also masks the original problem, which is why it took so long to notice. Does not affect Nexus mirrors (at least from the client; not sure what Nexus itself downloads), but does affect bits.netbeans.org. Forces us to publish a *.zip fallback index, which is a bigger download and does not support incremental downloads.
Since older IDE releases will not have the fix, still need to publish both formats. https://bitbucket.org/jglick/nexus-for-netbeans/changeset/eb8d43bab322 accomplishes that. (Currently we publish only the *.zip because of this bug. After the fix, an older IDE will actually download the *.gz index, fail to read it, then download the *.zip; oh well.)
(In reply to comment #1)
> core-main #62df07c4a7a6
Jesse, i agree with integration. Please do that before 10pm CET today, so we will ask NetCAT to test it during the weekend. Thanks in advance.
releases #92b966499c63 though after your deadline.
Integrated into 'main-golden', will be available in build *201103250400* on http://bits.netbeans.org/dev/nightly/ (upload may still be in progress)
User: Jesse Glick <email@example.com>
Log: #197039: V1-format (*.gz) Nexus indices cannot be fetched from Apache httpd
If I understand correctly, http://bits.netbeans.org/maven2/ now supports *.gz index, and ide after core-main #62df07c4a7a6 knows how to handle it (with fallback to *.zip)?
For main maven2 repo has nothing changed?
it seems like it's working properly, i was able to obtain index from all default repositories.
Product Version: NetBeans IDE 7.0 RC1 (Build 201103280000)
Java: 1.6.0_24; Java HotSpot(TM) 64-Bit Server VM 19.1-b02-334
System: Mac OS X version 10.6.7 running on x86_64; MacRoman; en_US (nb)
(In reply to comment #5)
> If I understand correctly, http://bits.netbeans.org/maven2/ now supports *.gz
> and ide after core-main #62df07c4a7a6 knows how to handle it
The IDE has long been able to handle *.gz indices (Central publishes one), but it was served incorrectly from our Apache server. With this fix, the IDE should be able to retrieve even the misserved *.gz index.
Additionally, we found the misconfiguration in the server and corrected it, so previous IDE versions should now be able to use the new-style index too.
> (with fallback to *.zip)?
In case *.gz is missing, the Indexer falls back to *.zip. Unfortunately this logic was broken in some older versions of the Indexer, including that used in Beta 2 IIRC, causing it to fail utterly if the *.gz was misserved even if the *.zip was present and valid.
(Sorry this is so complicated to verify, but the problem was in the interaction between client and server.)
> For main maven2 repo has nothing changed?
Not that I know of.
Update: MINDEXER-12 and MINDEXER-13 accepted upstream and should be in Maven Indexer 4.1.0; also related MINDEXER-20 (robustness against bad input) was applied.