On 2019-07-23 19:01, David L. Craig wrote:

I guess I need to ask frinnst et al. Running revdep has become de rigour to maintain CRUX platforms, it's just a question of how often. It seems to me revdep after every package install/update followed by rebuilding all flagged packages is becoming necessary. I would consider the platform broken as long as the affected packages are not rebuilt. But the impact of rebuilding qt5 etc. any time revdep determines the need is not trivial.

<--snip-->

So, everyone, what is best CRUX practice regarding revdep, and what changes, if any, do you think are indicated?

You generally (in a perfect world) would not need to run revdep. But as is the case with a source-based distribution and port maintainer screwups it will happen. Obviously I'm not saying that every time something breaks someone screwed up :-) We generally try to keep abi compabilities and library breakage down to an absolute minimum. Every time an expected breakage occurs, a notification should have been sent to this mailinglist. If you keep your CRUX installation up to date and is subscribed to this mailinglist it should be pretty easy to keep on top of. Such notifications usually look something like this: https://lists.crux.nu/pipermail/crux/2018-October/005939.html Obviously this gets a bit more tricky with new installations and the "first sysup" since that becomes a bit more unpredictable. As for recommendations or "best practice" it's hard to say. Obviously it's one of the first debugging steps to take if you encounter build errors or crashes. If you uninstall a library, a revdep run would be encouraged since even ports that doesn't *depend* on said library might link against it. So. What could be done to prevent these types of issues? Two things: 1. Create a policy that prevents port maintainers to break abi. That would almost make CRUX a source based debian-stable. 2. Prevent users from removing packages. Don't think any of those two choices are desirable ;-) /f