- Queries
- All Stories
- Search
- Advanced Search
- Transactions
- Transaction Logs
Advanced Search
Jan 18 2021
Jan 17 2021
Jan 14 2021
Jan 13 2021
Yeah, new ports can't have ports@ as maintainer, and I think that would apply by extension to restored ones.
Jan 12 2021
Jan 11 2021
Jan 8 2021
The plist has gotten big enough that you should move it to a separate pkg-plist file.
Jan 5 2021
In D27954#624798, @grembo wrote:In D27954#624793, @swills wrote:It seems to me, and maybe I'm wrong, that this particular case could have been avoided by checking the output of make makeplist more carefully, it's not perfect, never has been and never will be. Am I missing something?
This would have avoided installing a couple of .bak files by accident (by the way, pkg-plist is 9000+ lines in this case), which wouldn't have any impact on the package, or its functionality (or even really the size of the package). As this port never created .bak files in the past (due to REINPLACE_ARGS being -i ""), I also didn't specifically look for them. I relied on the framework's QA steps to catch plist problems (the kind that breaks a build, not that kind that might install an unnecessary 1kb file by accident).
To me, the usual process is:
- Do update (make, make makesum etc.)
- Run make makeplist
- Do a lot of testing if the port actually still works as expected.
- Visually inspect plist and fix the usual suspects (like rc scripts generated from in files).
- Run various qa targets manually (with DEVELOPER in make.conf).
- Run on multiple poudriere instances using poudriere testport
I can, of course, add poudriere bulk to this procedure, but my basic assumption would be that if something passes all checks in DEVELOPER mode, it should also build outside of DEVELOPER mode.
I don't really understand why you oppose a simple fix that has no negative side effects, makes broken things entering the ports tree less likely, and can be removed easily in case you get rid of REINPLACE_ARGS in the future. I also offered to do something more elaborate (like, still do all the things that sed_checked does, without leaving files behind), but I assumed something very simple and easy to audit would be preferable as a stop-gap solution. Instead I'm told that "all I want to do is the stop the framework telling me they [my port] are broken" and that it's my own fault for relying on the framework/not proof-reading pkg-plist carefully enough.
It seems to me, and maybe I'm wrong, that this particular case could have been avoided by checking the output of make makeplist more carefully, it's not perfect, never has been and never will be. Am I missing something?
I agree with @mat, it's probably better to avoid the problem by not allow overriding REINPLACE_ARGS in the first place. There aren't that many ports that override it, looks to be less than 90, so not that big of a clean up IMHO.