Re: flex is no longer M-A:foreign
Hi Manoj,
On Sun, Feb 14, 2016 at 10:03:17PM -0800, Manoj Srivastava wrote:
> So, we have between 5 and 6 packages in this list of 3000. I???ll
> see about preparing patches to submit for these, and I think we should
> be able to just move the current version in experimental (the ohne that
> does not pull in libfl-dev) to unstable, since the number of packages
> affected seems low.
>
> Have I missed something here?
Yes, you are missing something. Did you encounter src:mig thus far? It
also fails when libfl-dev.
So those 5 packages I mentioned are "true positives". My mail also
contained a larger number of "false positives" and what you didn't
account for is an unknown number of "false negatives". Those cross
builds, I was grepping there covered 3000 source packages. Rounding
down, that'd be 10% of the archive. Assuming that those 3000 source
packages are probabilistically independent of the 500 packages using
flex, we'd expect 50 relevant ones to be cross built. That assumption of
course is optimistic as well (notably none of the gccs were cross
built). Of those 50, more than half failed for unrelated reasons. Using
this optimistic gauge, I'd thus expect that 5/25 packages need changes.
My (optimistic) guess thus is that we break 100 source packages.
What we need is better data. Unfortunately, I cannot provide it atm.
Nevertheless, deciding which route to take (dropping the flex ->
libfl-dev dependency or renaming flex to flex-bin) is beneficial right
now. It is the premise for filing bugs in reverse dependencies and
fixing them and we certainly should give them a reasonable time to fix.
Do I understand correctly that you are indeed going to drop the flex ->
libfl-dev dependency? Can I base bug+patch reports on that plan?
Helmut
Reply to: