If we don't make such a list, we will end up having multiple tickets, that will never start coming with requests to install same packages
On Thu, Nov 7, 2013 at 4:39 PM, Petr Bena <[email protected]> wrote: > I am not talking about "list of installed packages" I am talking about > "list of packages that were already requested by other users, are in > puppet and guaranteed to be present, so that you don't need to bother > us with requesting them again, because they were already requested and > are installed" > > On Thu, Nov 7, 2013 at 4:35 PM, Marc A. Pelletier <[email protected]> wrote: >> On 11/07/2013 10:26 AM, Petr Bena wrote: >>> From what Tim said, it seems that I should fill in a bug report for >>> every single package no matter if it's already in puppet or not, >>> because there is clearly no simple way for me to verify that. >> >> Yes, you should, because otherwise you're relying on implicit >> dependencies, which is iffy at best. If you have a tool that needs >> libfoo-perl, then you need to make sure libfoo-perl is puppetized and >> not rely on the fact that foobar-tools happens to have it as a >> dependency and it got installed as a consequence. >> >> OTOH, never request installation of dependencies of things you depend >> on; rely on apt to work those out itself. >> >> In other words, if your tool invokes 'gizmo' from the package 'gizmo', >> list *that* as a dependency, and not the 'libgizmo' it depends on in >> turn - you never know when gizmo will switch to libwidget instead. >> >> This is why making or comparing a "list of installed packages" is >> exactly the wrong way to go about things -- there might be things >> leftover on nodes from previous versions, or things that were installed >> but then never removed (changes in puppet, in general, do not purge >> packages no longer explicitly installed). >> >> -- Marc >> >> >> _______________________________________________ >> Labs-l mailing list >> [email protected] >> https://lists.wikimedia.org/mailman/listinfo/labs-l _______________________________________________ Labs-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/labs-l
