> On May 8, 2025, at 11:28, Helge Oldach <[email protected]> wrote: > > Moin Rahman wrote on Thu, 08 May 2025 10:14:37 +0200 (CEST): >> This is not about being hostile to `make install` or `portmaster` users. It's >> about scope and sustainability. The FreeBSD ports tree is under continuous >> development. If you're filing bugs against outdated snapshots - with stale >> options, outdated configs, or assumptions that haven't held for multiple >> quarterly cycles - then you're not reporting problems in supported code. >> You're >> asking others to chase ghosts. > > True, however at the end of the day, poudriere is also running said > `make install` or a procedure similar to `portmaster`. Therefore it is > very sensible to seriously consider related bug reports - provided they > are against an up-to-date tree of course. > > Secondly, building ports directly provides additional benefits, such as > selecting specific OPTIONS. This is not available through pre-built pkg. > There is a reason why OPTIONS are there, so reports against them should > be taken seriously rather than brushed off in favor of pkg/poudriere's > one-size-fits-all option choice. > >> Poudriere exists precisely to avoid these problems: it gives you clean, >> reproducible builds that reflect the real state of the tree - not whatever >> cruft >> is left behind in /var/db/ports or custom local configs or polluted >> environment. >> If you aren't using poudriere and aren't tracking HEAD or a supported >> quarterly >> snapshot, you're on your own. That's the reality - not out of hostility, but >> necessity. > > Is it also responsible for this? > > # pkg install rust > Updating FreeBSD repository catalogue... > Fetching meta.conf: 100% 179 B 0.2kB/s 00:01 > Fetching data.pkg: 100% 10 MiB 2.6MB/s 00:04 > Processing entries: 100% > FreeBSD repository update completed. 35918 packages processed. > All repositories are up to date. > pkg: No packages available to install matching 'rust' have been found in the > repositories > # > > To be clear: `make -C /usr/ports/lang/rust install` works fine (but > takes ages). And `pkg install rust-nightly` is fine as well. > > Going back to the question raised in the subject line, I would expect > that bug reports against specific port builds are being well considered, > as they are the basis for poudriere producing the right results in > general. > > Kind regards > Helge
Hi Helge, Your reply misrepresents both poudriere and the point I made earlier, so let’s correct that before further confusion spreads. First, poudriere fully supports building ports with custom OPTIONS. It’s not locked to defaults — it lets users define and persist OPTIONS cleanly, per jail or build list, with complete control. If you believe poudriere users can’t build with specific OPTIONS, then I’m afraid you haven’t used it seriously. That makes your comment — in the middle of a thread focused on the reliability of reproducible builds — not particularly helpful. Second, the pkg install rust example you brought up is a red herring. That failure was due to a temporary issue in the latest repository on FreeBSD 14.X — not a problem with poudriere, and not relevant to the topic at hand. If you’re using latest, you're accepting some level of churn. If that’s unacceptable, use quarterly. That’s what it exists for. Finally, my recommendation to use poudriere had nothing to do with using default package options. It was about building your own packages with your own options — but in a way that’s reproducible and supportable. That’s how you get reliable behavior, avoid stale /var/db/ports artifacts, and report bugs that developers can actually triage. We need to be clear and technically correct when advising users. If someone wants to use manual make installworkflows, that’s their prerogative — but bug reports need to reflect a supported tree, current behavior, and reproducible build state. That’s why poudriere matters — not because it runs make, but because it does so correctly. Kind regards, Moin
signature.asc
Description: Message signed with OpenPGP
