ARM development: next steps

[davidep (Davide Principi)]

@mark_nl

(1) most because rspamd 1.8.2 does not play along: it won’t compile, 1.8.1 and 1.8.3 does

@stephdl can we go with 1.8.3 for x86_64 too?

[davidep (Davide Principi)]

@mark_nl

just a few conditionals in /usr/sbin/nethserver-install , athough it does not seem to be impossible to workaround this in the end.

Conditionals could be acceptable, but I’m not sure they’re really required. Instead, I think the mirrorlist nethserver.php script requires some changes for additional architectures.

[mark_nl (Mark Verlinde)]

@davidep

A noarch package would be cloned to x86_64, armhfp and aarch64 ¹, whilst for arch-dependant packages the rule is

• x86_64 goes to x86_64 repo
• armv7hl goes to armhfp repo
• aarch64 goes to aarch64 repo

As the upload permissions are quite limited, when uploading an ARM package the repository name has to be prefixed with arm- . For example:

• arm-base
• arm-updates
• arm-testing
• arm-nethforge
• arm-nethforge-testing

Agreed, works for me!

@davidep

do we need a blacklist to inihibit the automatic copy? IIRC we said: yes

Yes some modules need to be blacklisted / or greylisted:

• not all packages are ported, mainly those depending on a kmod package
• some modules may turn out not to be feasible, this could be controlled with the comps file to.: meaning they are in the repo for those who want to install them but wont show up in the software-center
• if a module has newer dependencies (rspamd as an example) they need to be held back until this dependency is in place. IIRC this works automatic with a repo closure method.

@davidep

(sorry if already answered) do we need our arm-epel or just forward to the “official” Index of /repodir/epel-pass-1 ?

For simplicity I tend to be in favor to drop the (ns) epel mirror. Until now no update did actually cause to break something. Moreover maintenance of the repository has improved lately.

However the packages will remain unsigned and updates are released untested.

To keep momentum we can opt to omit the epel mirror in this stage, reaching in epel-pass-1 directly and decide how to handle this later (IIRC @dz00te suggesting this all along) .

@davidep

I hope to finish my task on packages today. If you’re blocked by my work proceed with the current repository!

I will proceed with some early testing; however my preference is to release the 7.6 alpha image based on the final repositories with corresponding packages for nethserver-realease and nethserver-mock.

[stephdl (Stéphane de Labrusse)]

@davidep

@stephdl can we go with 1.8.3 for x86_64 too?

we know a little issue, the history tab (in the UI is filled by the learn spam errors), I need to ask upstream.

This is just cosmetic, not blocking for the spam filtering

[rgmhtt (Robert Moskowitz)]

I now have my Odroid HC1 ready for testing. If I have figured this out right, the boot partition (and all the bootloader and uboot) cruft on the uSD card can stay in place, and I can just reimage the sata drive and test and reimage…

Is there a ‘simple’ way for me to get a working Nethserver going on it? Obviously the first thing will be to work it up as a mail server. As much as I need to replace my ClearOS PDC, I need to upgrade my public-facing (and public accessed) mail server.

I run a split-zone DNS (External and Internal) and for starters, I can do all my testing internally before I let it out loose on wild.

[mark_nl (Mark Verlinde)]

@rgmhtt

Is there a ‘simple’ way for me to get a working Nethserver going on it?

Sorry, at the moment it is a bit stale because of the update to 7.6.1810.
I hope we have an offer end of the week.

[mark_nl (Mark Verlinde)]

@stephdl

@stephdl can we go with 1.8.3 for x86_64 too?

we know a little issue, the history tab (in the UI is filled by the learn spam errors), I need to ask upstream.

This is just cosmetic, not blocking for the spam filtering

Found the issue why rspamd 1.8.2 does not compile with gcc 4.8.5. i’ll patch it to build.

fix preprocessor syntax

[stephdl (Stéphane de Labrusse)]

1.8.2 gets the same issue, will try today to ask upstream

@davidep we should fix on rspamd-1.8.1 if it is not too late for NS7.6

[giacomo (Giacomo Sanchietti)]

@stephdl

we should fix on rspamd-1.8.1 if it is not too late for NS7.6

Too late

I don’t see any problem on using 1.8.1 or 1.8.3 for ARM.
Binary packages will differ for sure even for other applications.

[davidep (Davide Principi)]

@mark_nl

Index of /repodir/epel-pass-1

I’m setting up the repoclosure configurations. I’d need an rsync mirror

Do you know if it (already) exists?

[mark_nl (Mark Verlinde)]

@davidep

I’d need an rsync mirror

As said before I do not know much (understatement) about repo mirror’s.
Can you try to explain what a rsync mirror is?

[davidep (Davide Principi)]

Sorry for being too generic!

I’d need an rsync mirror of armhfp EPEL7. I couldn’t find one and I’m considering to fetch required data via http. I just need repodata/ contents to run repoclosure…

[mark_nl (Mark Verlinde)]

https://armv7.dev.centos.org/repodir/epel-pass-1/repodata/

This?

[stephdl (Stéphane de Labrusse)]

yes…rspamd-1.8.3 brings some UI enhancements, it is good too

[mark_nl (Mark Verlinde)]

@davidep

As the upload permissions are quite limited, when uploading an ARM package the repository name has to be prefixed with arm- . For example:

• arm-base
• arm-updates

It’s exiting seeing it grow and become factual

Do I understand correctly the arm packges are uploaded (as example) to arm-base/armhfp/Packages and installed form base/armhfp/Packages ?

[giacomo (Giacomo Sanchietti)]

@mark_nl

Do I understand correctly the arm packges are uploaded (as example) to arm-base/armhfp/Packages and installed form base/armhfp/Packages ?

I’d say yes … let’s see if @davidep confirm if we both correctly understood how it work

[davidep (Davide Principi)]

I hope the new repositories work well. I’d expect some bugs, however I have to complete the mirrorlist fix now. I’d like to avoid forking the nethserver-release package and retain it as arch-independent.

@mark_nl

To keep momentum we can opt to omit the epel mirror in this stage

Ok, I will handle it from mirrorlist.

[davidep (Davide Principi)]

@davidep

Ok, I will handle it from mirrorlist.

Ops! I forgot epel is not handled by mirrorlist.nethserver.org

I guess an ARM device must override epel-release by other means. For instance the basic image could provide a customized /etc/yum.repos.d/epel.repo.

[mark_nl (Mark Verlinde)]

@davidep

Ops! I forgot epel is not handled by mirrorlist.nethserver.org

Sorry to be unclear

@mark_nl

a package which provides “epel” for arm32

I meant a package containing repo-settings. in /etc/yum.repos.d/ for “epel”.