On May 28, 2016 4:20 AM, "Dmitry Smirnov" <only...@debian.org> wrote: > > On Friday, 20 May 2016 12:50:48 AM AEST Austin English wrote: > > I understand that there's a *potential* problem. > > Potential problem of breaking APT operations is much more dangerous than > "actual" problem of ignoring return values. > > > > But > > here we have an *actual* problem, where the script is silently ignoring > > errors which is causing problems in downstream, and as far as I can > > tell, the current solution is to ignore actual errors in favor of > > preventing *potential* solutions? > > That's right. Script was not intended to be used directly to begin with. > It was designed to be run automatically at the end of APT operations and > (re-)build passenger package when required. > > > > What exactly needs to be tested in the current packaging scheme so that > > behavior could be changed? I volunteer to help with that, if it's made > > clear what needs to be done. > > Perhaps to temporary break script to always return error value and try > various apt operations (e.g. install, remove and reconfigure of other > packages) when post-invoke hook is activated to perform build or update. > > We need to be absolutely sure that build/install errors do not propagate > where they does not belong and that nothing got broken if we take > responsibility for propagation of those errors. > > -- > Best wishes, > Dmitry Smirnov. > > --- > > Without doubt you are not sane. > -- Tage Danielsson
Thanks for reply Dmitry, I'll put this on my backlog.
