apport-hackers team mailing list archive

[Brian Murray <brian@xxxxxxxxxx>]

> With this implementation, if apport-retrace is called with --gdb-sandbox, no
> dbg packages are installed for the extra packages (because
> extra_packages==gdb); but if called with --gdb-sandbox --extra-packages=foo,
> both foo-dbg and gdb-dbg will be installed.  This seems inconsistent, but
> because we're using the install_dbg argument to
> apport.packaging.install_packages(), we're all or nothing.
> 
> I guess what I would ask here is, do we have reason to consider it correct to
> automatically install the dbg packages along with extra packages?  

No, I was trying to main consistency with the previous behavior but agree that if people want the dbg packages they should be passed via extra packages.

> There may
> or may not be existing users who rely on this behavior, but *should* they?
> Should they not instead provide a complete list of the extra packages that
> they want installed, including any debug packages?
> 
> I would suggest one of two options here:
> 
> - if there's no requirement to automatically install dbg packages with extra-
> packages, just unconditionally pass install_dbg=False.

I'll go with this option.

> - if there is a requirement to automatically install dbg packages for extra-
> packages, just call install_packages() separately for gdb vs. the user-
> specified extra_packages.  (We already call install_packages() twice
> currently, a third call should hopefully be ok?)

The install_packages calls actually do some work building the sandbox and I'd prefer to figure out how to make it one call if possible.
-- 
https://code.launchpad.net/~brian-murray/apport/speedier-sandboxes/+merge/335681
Your team Apport upstream developers is requested to review the proposed merge of lp:~brian-murray/apport/speedier-sandboxes into lp:apport.