We should spend a bit of time in the retrospective talking about how
the platform is consuming bits from other projects. There were some
topics around ECF and this one around p2. Both are from another top
level project. Clarifying the rampdown and approvals process there
would be useful to eliminate confusion about who needed to approve what
etc.
Jeff
Mike Wilson wrote:
*sigh*. Yes, I'm tired.
How about "Obviously, whatever can be done to bound the
impact of this code would be very valuable."
McQ.
Mike
Wilson---11/06/2009 12:30:34---For something as critical as p2, I'd
argue that we need to fix any regression that some consumer has already
hit. It's almost c

From: |

Mike Wilson/Ottawa/IBM@IBMCA |

To: |

eclipse-pmc@xxxxxxxxxxx |

Date: |

11/06/2009 12:30 |

Subject: |

Re: [eclipse-pmc] Request to fix 279542 |
For something as critical as p2, I'd argue that we
need to fix any regression that some consumer has already hit. It's
almost certain that there will be others who will see this as they move
from R3.4 to R3.5.
I've looked at the patch, and as has been noted, it's bigger/more
significant than we would like to be making, at this point. Obviously,
whatever can be done to bound the impact of this code. For example,
does it only run in cases that would have failed anyway, without the
fix?
McQ.
Pascal
Rapicault---11/06/2009 12:14:44---I'm soliciting the input of the PMC
in order to address bug https://bugs.eclipse.org/bugs/show_bug.cgi?id=279542

From: |

Pascal Rapicault/Ottawa/IBM@IBMCA |

To: |

eclipse-pmc@xxxxxxxxxxx |

Date: |

11/06/2009 12:14 |

Subject: |

[eclipse-pmc] Request to fix 279542 |
I'm soliciting the input of the PMC in order to address bug https://bugs.eclipse.org/bugs/show_bug.cgi?id=279542
Background:
The TPTP team has recently discovered that the installation through p2
of their agent controller functionality is not working.
The root cause of it is the failure of the p2 ln action (to create
symbolic links) when passed a special token (in this case @artifact).
The failure is not dramatic since it does not fail the installation,
but it leaves the agent controller in a non-working state since links
are missing. This is regression from 3.4.2 and the functionality they
were using was "documented".
So far they are the only team that have complained about this problem,
but it is unclear if others from the community are also affected.
We have a patch ready in which we have a good confidence but it is not
a one liner, but it has been tested by the TPTP team. This patch reused
some of the 3.4.2 code plus some others pieces of p2 code we had.
The risk:
Worst case is we break other install scenarios using other links
facilities.
Reward:
We are fixing a regression. Potentially fixing other people. TPTP has
nothing to do.
Workaround:
We have been in touch with the TPTP team and have explored several
possibilities. One involving no p2 change consists in having them
adding some code in a plug-in activator to create the missing links.
This solution is "simple" but requires probably as much code for them
and it would also imply some build changes since their plug-in does not
have java code.
Also whatever workaround they would ship with would have to be kept in
3.5.1 since the update will run with 3.5.0 code.
PaScaL
_______________________________________________
eclipse-pmc mailing list
eclipse-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
_______________________________________________
eclipse-pmc mailing list
eclipse-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
_______________________________________________
eclipse-pmc mailing list
eclipse-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
|