Red Hat Bugzilla – Bug 1268322
vector bundling broken in scripts/spin-rawhide
Last modified: 2015-10-26 20:21:21 EDT
Upstream pcp commit 81f4ac46ff7 broke bundling of vector in pcp. An examination of the pcp-webapp-vector-*git*.noarch.rpm files shows source checkouts of the vector repository, not a deployable or even runnable version. Directing a web browser at http://localhost:44323/vector confirms the breakage. See also http://oss.sgi.com/pipermail/pcp/2015-September/008310.html .
Build scripts fixed upstream last week by myself & lberk - can you close out with this weeks rawhide spin please, Lukas? Taa.
It would be nice if the rawhide snapshot of vector contained some version information. The current code obliterates the webapps/index.html version info, but should instead transcribe an identifier (e.g. git-describe) of the vector snapshot.
> It would be nice if the rawhide snapshot of vector contained some version information. Ideally this info would be made visible in the UI somewhere, where any/all users would be able to report it to devs (thinking beyond just rpm users and rawhide) - I'll mention this to Martin. > The current code obliterates the webapps/index.html version info, but should instead transcribe an identifier (e.g. git-describe) of the vector snapshot. This (top-level pcp-webjs index.html) was always the wrong place to put version information - there is no requirement to install both pcp-webjs and pcp-webapp-vector, and nor should there be. In reality, "obliterates" is just "fixing the bug".
> Ideally this info would be made visible in the UI somewhere, where any/all > users would be able to report it to devs (thinking beyond just rpm users and > rawhide) - I'll mention this to Martin. Since you are taking arbitrary git snapshots, the build process would have to include git version info into the dist/ parts, rather than have a simple "release X.Y" hardcoded string. > > The current code obliterates the webapps/index.html version info, but should instead transcribe an identifier (e.g. git-describe) of the vector snapshot. > > This (top-level pcp-webjs index.html) was always the wrong place to put > version information - there is no requirement to install both pcp-webjs and > pcp-webapp-vector, and nor should there be. > > In reality, "obliterates" is just "fixing the bug". There is no "wrong ... bug". The information in the index.html may become redundant (once vector learns to self-report) or unavailable (someone not installing pcp-webjs), but not incorrect ("bug"?). But for the moment, it is the only game in town.
(In reply to Frank Ch. Eigler from comment #4) > [...] > There is no "wrong ... bug". The information in the index.html may > become redundant (once vector learns to self-report) or unavailable > (someone not installing pcp-webjs), but not incorrect ("bug"?). It really was incorrect. One can install a more recent or older pcp-webapp-vector RPM alongside the (totally unrelated) pcp-webjs RPM. The pcp-webjs index.html file would then have happily reported an incorrect version number for Vector. But that was fixed awhile back in the spec, so not really worth continued discussion here.
> One can install a more recent or older pcp-webapp-vector RPM alongside the > (totally unrelated) pcp-webjs RPM. The pcp-webjs index.html file would then > have happily reported an incorrect version number for Vector. > > But that was fixed awhile back in the spec, so not really worth continued > discussion here. If that occurred, that would have been a self-made problem caused by the unnecessary and imperfect subdivision of the webjs suite. That says nothing about the propriety of providing information in the index.html file.
Fixed for some time. Chatted to Martin last week, and he has kindly offered to add a mechanism for exposing version information via the Vector UI at some point.