Обсуждение: Proposal for "common" repository
Hi, (First of all: This is not an April 1st thing) Currently we have two types of RPMs: Some of them are tied to a specific PostgreSQL version, and some of them are not (I will call them "common"). These common packages bring in two problems: * The current build infrastructure builds them again and again for all distros+testing repos. For example, Proj is built on v12, v12testing, v11, v11testing, v10, v10testing, v9.6, v9.6testing, and v9.5 repositories. (I'll probably stop 9.6 and 10 testing repos soon, but that is not the topic right now.) * 1 common package is kept in 9 different directories (see above), which consumes a lot of disk space. It generates lots of FTP and rsync traffic. So, to avoid these two issues, I'm inclined to make a few changes: Git repo: - Rename the current "common" directory (which handles Makefiles) to "global" directory (to avoid confusion) - Make sure that common and "non-common" packages are in separate directories in git, so I'll move many packages to a new subdir - Adjust makefiles in global, and invent new directories for those common RPMs (Should be easy by altering _srcrpmdir and _rpmdir variables there), and new targets for those common RPMs. Build servers: - Update build *and sync* scripts for the new layout - Update repo files for the new layout So, this will definitely break lots of things, but I feel like I have to bite this bullet anyway. Any comments, objections? Regards, -- Devrim Gündüz Open Source Solution Architect, Red Hat Certified Engineer Twitter: @DevrimGunduz , @DevrimGunduzTR
Вложения
On Wed, 2020-04-01 at 12:12 +0100, Devrim Gündüz wrote: > So, to avoid these two issues, I'm inclined to make a few changes: > > Git repo: > > - Rename the current "common" directory (which handles Makefiles) to "global" > directory (to avoid confusion) > > - Make sure that common and "non-common" packages are in separate directories > in git, so I'll move many packages to a new subdir > > - Adjust makefiles in global, and invent new directories for those common RPMs > (Should be easy by altering _srcrpmdir and _rpmdir variables there), and new > targets for those common RPMs. > > So, this will definitely break lots of things, but I feel like I have to bite > this bullet anyway. > > Any comments, objections? Will this break regular updates like "yum upgrade"? If yes, then I am a bit doubtful. That would make many people unhappy and lead to people running outdated minor releases. Yours, Laurenz Albe
Hi, On Wed, 2020-04-01 at 13:20 +0200, Laurenz Albe wrote: > Will this break regular updates like "yum upgrade"? No, if they use the new repo RPM. This is my current plan: ================================================== # PGDG Fedora testing - common repository [PGDGCommonTestingRPM] name=PostgreSQL common RPMs for Fedora $releasever - $basearch Testing Repo baseurl=https://download.postgresql.org/pub/repos/yum/testing/common/fedora/fedora-$releasever-$basearch enabled=1 gpgcheck=1 gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-PGDG ================================================== > If yes, then I am a bit doubtful. > That would make many people unhappy and lead to people running outdated > minor releases. I'm open to ideas in here in order to minimize the breakage. Regards, -- Devrim Gündüz Open Source Solution Architect, Red Hat Certified Engineer Twitter: @DevrimGunduz , @DevrimGunduzTR
Вложения
Hi again, On Wed, 2020-04-01 at 12:40 +0100, Devrim Gündüz wrote: > No, if they use the new repo RPM. Just to clarify: Users will be able to get PostgreSQL updates, and many other "non-common" updates even if they don't update the repo files. However, (for example) PostGIS updates will fail, because they won't be able to get new Proj packages when I update them. At that point, they will have to update the repo file. So the major user base won't be affected. I'm more concerned about the folks who build their own packages from our repo. Their scripts will be broken. Regards, -- Devrim Gündüz Open Source Solution Architect, Red Hat Certified Engineer Twitter: @DevrimGunduz , @DevrimGunduzTR