[
https://issues.jboss.org/browse/JBIDE-19757?page=com.atlassian.jira.plugi...
]
Nick Boldt commented on JBIDE-19757:
------------------------------------
We keep the site.xml around because the old WTP server adapter discovery required it: the
old adapter couldn't parse regular p2 sites (or composite sites!) and needed the
site.xml to define which feature mapped to the required category name.
So, if your perf tests are accurate, then +1 for:
* trashing site.xml from the JBT aggregate site,
* removing the webtools site build,
* symlinkin from
http://download.jboss.org/jbosstools/updates/webtools/mars/ into
http://download.jboss.org/jbosstools/mars/development/updates/.
Did you test the new server adapter discovery against a *composite* site, rather than just
an aggregate site like
http://download.jboss.org/jbosstools/mars/snapshots/builds/jbosstools-bui...
?
If composites still don't work, we can instead symlink to
http://download.jboss.org/jbosstools/static/mars/development/updates/core... but
that will mean having to update the symlink every time there's a new dev milestone
release; I'd prefer (if possible) to just link to the latest available milestone (and
then we only need to change the pointer when .Final is released).
Consider using main JBT site for server discovery
-------------------------------------------------
Key: JBIDE-19757
URL:
https://issues.jboss.org/browse/JBIDE-19757
Project: Tools (JBoss Tools)
Issue Type: Enhancement
Components: build, server
Reporter: Mickael Istria
Fix For: 4.3.0.Beta1
With
https://bugs.eclipse.org/bugs/show_bug.cgi?id=434185 , WTP Server Discovery
mechanism was granted a new strategy which allows to rely on regular p2 metadata instead
of a site.xml.
Support for this was already merged in server (
https://github.com/jbosstools/jbosstools-server/commit/2d3cc63a9b67753ad9...
)
In order to save an artifact to manage (the webtools p2 repository), we could use this
mechanism and consider contributing directly the main JBT URL to webtools discovery.
However, server discovery also keeps older strategies and since we produce invalid
site.xml files, this is currently failing
{code}
!ENTRY org.eclipse.equinox.p2.updatesite 2 0 2015-05-04 09:40:58.088
!MESSAGE Error parsing feature stream. The unique identifier or the version is null or
empty for the State: "Category": unique identifier="minimal-json"
version="null".
{code}
because we are lines specifying bundle but no version in the site.xml.
[~nickboldt] What are those site.xml useful for? Could we get rid of them?
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)