https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >= 1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
Rich.
On Tue, Mar 1, 2022 at 3:07 AM Richard W.M. Jones rjones@redhat.com wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >= 1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
It's not crazy. EPEL is intended to build on RHEL content, which means we can't depend on something RHEL doesn't publish. If Red Hat wants to publish their buildroot repo, then sure, we could use it. Just because it happens to exist in the CentOS Stream 9 buildroot content does not mean we would be able to rely on it once we replace CentOS Stream with RHEL for EPEL 9. Thus, we don't use the CentOS Stream 9 buildroot either.
If we did, we'd wind up in a situation where packages were built once and then not buildable ever again. That already kind of happened when we initially had that buildroot repo in the EPEL build environment and it made it way harder for us to figure out what gaps we had for things to build against RHEL later. We've fortunately dealt with the small number of cases that occurred from then.
-- 真実はいつも一つ!/ Always, there's only one truth!
On Tue, Mar 01, 2022 at 04:21:56AM -0500, Neal Gompa wrote:
On Tue, Mar 1, 2022 at 3:07 AM Richard W.M. Jones rjones@redhat.com wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >= 1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
It's not crazy. EPEL is intended to build on RHEL content, which means we can't depend on something RHEL doesn't publish. If Red Hat wants to publish their buildroot repo, then sure, we could use it.
I wasn't very clear, but I was addressing my remark at Red Hat. There's really no reason why we (Red Hat) don't publish buildroot, in fact my personal view is we ought to for open source reasons.
Just because it happens to exist in the CentOS Stream 9 buildroot content does not mean we would be able to rely on it once we replace CentOS Stream with RHEL for EPEL 9. Thus, we don't use the CentOS Stream 9 buildroot either.
So this was going to be my next question - is it that difficult to use C9S buildroot packages to replace the "missing" ones? AFAIK they ought to be almost identical. Obviously they are rebuilds and they might be a little out of sync, but saves EPEL doing a literal third rebuild of the same content!
If we did, we'd wind up in a situation where packages were built once and then not buildable ever again. That already kind of happened when we initially had that buildroot repo in the EPEL build environment and it made it way harder for us to figure out what gaps we had for things to build against RHEL later. We've fortunately dealt with the small number of cases that occurred from then.
I'm not sure I totally understand this bit. Is it right to say that packages wouldn't be "buildable ever again" only in the case where we used C9S buildroot and then dropped it? If we just use C9S buildroot packages + RHEL 9 packages - forever - we'd be OK?
Rich.
On Tue, Mar 01, 2022 at 01:19:32PM +0000, Richard W.M. Jones wrote:
So this was going to be my next question - is it that difficult to use C9S buildroot packages to replace the "missing" ones? AFAIK they ought to be almost identical. Obviously they are rebuilds and they might be a little out of sync, but saves EPEL doing a literal third rebuild of the same content!
But to add - ideally we wouldn't need to do this if Red Hat published the buildroot packages, and I don't understand why we don't do this.
Rich.
On Tue, Mar 1, 2022 at 8:20 AM Richard W.M. Jones rjones@redhat.com wrote:
On Tue, Mar 01, 2022 at 04:21:56AM -0500, Neal Gompa wrote:
On Tue, Mar 1, 2022 at 3:07 AM Richard W.M. Jones rjones@redhat.com wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >= 1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
It's not crazy. EPEL is intended to build on RHEL content, which means we can't depend on something RHEL doesn't publish. If Red Hat wants to publish their buildroot repo, then sure, we could use it.
I wasn't very clear, but I was addressing my remark at Red Hat. There's really no reason why we (Red Hat) don't publish buildroot, in fact my personal view is we ought to for open source reasons.
Just because it happens to exist in the CentOS Stream 9 buildroot content does not mean we would be able to rely on it once we replace CentOS Stream with RHEL for EPEL 9. Thus, we don't use the CentOS Stream 9 buildroot either.
So this was going to be my next question - is it that difficult to use C9S buildroot packages to replace the "missing" ones? AFAIK they ought to be almost identical. Obviously they are rebuilds and they might be a little out of sync, but saves EPEL doing a literal third rebuild of the same content!
Theoretically, yes. And for some stuff, that would work. It depends on how sensitive things are and where they lie in the dependency chain.
If we did, we'd wind up in a situation where packages were built once and then not buildable ever again. That already kind of happened when we initially had that buildroot repo in the EPEL build environment and it made it way harder for us to figure out what gaps we had for things to build against RHEL later. We've fortunately dealt with the small number of cases that occurred from then.
I'm not sure I totally understand this bit. Is it right to say that packages wouldn't be "buildable ever again" only in the case where we used C9S buildroot and then dropped it? If we just use C9S buildroot packages + RHEL 9 packages - forever - we'd be OK?
Those packages are not necessarily guaranteed to be installable forever because they're effectively only there as a side-effect. But it's *possible* we'd be fine.
There is another issue with using buildroot packages: they're not signed and mirrored at all. There's no reasonable way to expect downstreams to be able to figure out how to build our packages with any reasonable trust. People already don't like the fact that RHEL doesn't do it, I think they'd be extremely upset if it led to EPEL packages that people couldn't easily locally (re)build and update. It's also more common for EPEL packagers to be in environments where unsigned content is simply flat out blocked by policy, so RPM would trip up over installing those packages.
And yay supply chain attacks! :(
-- 真実はいつも一つ!/ Always, there's only one truth!
On Tue, 1 Mar 2022 at 08:19, Richard W.M. Jones rjones@redhat.com wrote:
On Tue, Mar 01, 2022 at 04:21:56AM -0500, Neal Gompa wrote:
On Tue, Mar 1, 2022 at 3:07 AM Richard W.M. Jones rjones@redhat.com
wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >=
1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
It's not crazy. EPEL is intended to build on RHEL content, which means we can't depend on something RHEL doesn't publish. If Red Hat wants to publish their buildroot repo, then sure, we could use it.
I wasn't very clear, but I was addressing my remark at Red Hat. There's really no reason why we (Red Hat) don't publish buildroot, in fact my personal view is we ought to for open source reasons.
I do not think you will find much disagreement here.. but after 3+ years of saying it and nothing changing, many of us have made our peace.
Just because it happens to exist in the CentOS Stream 9 buildroot content does not mean we would be able to rely on it once we replace CentOS Stream with RHEL for EPEL 9. Thus, we don't use the CentOS Stream 9 buildroot either.
So this was going to be my next question - is it that difficult to use C9S buildroot packages to replace the "missing" ones? AFAIK they ought to be almost identical. Obviously they are rebuilds and they might be a little out of sync, but saves EPEL doing a literal third rebuild of the same content!
The issue in the past has been that it takes manual matching at times to make it work. Koji, mock and rpmbuild will all complain in different ways when package content varies in minute ways. Someone then has to rebuild that package when that happens, which usually only is found at 2am by some very cranky engineer who posts a lot of less than polite messages about how EPEL people are complete crap. Or you find that the internal package is good enough to have built the RHEL content but still is lacking something that you expected to be there for anything NOT a RHEL content. Again that takes inspection and someone to care enough to do it. If we need that, then we might as well rebuild the content and make sure it is what we wanted in the first place.
If we did, we'd wind up in a situation where packages were built once and then not buildable ever again. That already kind of happened when we initially had that buildroot repo in the EPEL build environment and it made it way harder for us to figure out what gaps we had for things to build against RHEL later. We've fortunately dealt with the small number of cases that occurred from then.
I'm not sure I totally understand this bit. Is it right to say that packages wouldn't be "buildable ever again" only in the case where we used C9S buildroot and then dropped it? If we just use C9S buildroot packages + RHEL 9 packages - forever - we'd be OK?
The way Fedora build system deals with RHEL packages is a bit of 'hack' compared to how it builds Fedora packages. It sees them as external packages and (mis)-uses a method which was originally only for bootstrapping a distro to see packages it did not build itself. This then requires additional hacks on top of that to keep the facade working.. those hacks break regularly and have to be manually dealt with. Usually the breakage then requires some 'we can break the koji database again so do a full backup and possibly a restore' actions by Fedora release engineering to delete some entry koji 'thinks' should be there but isn't (or vice versa). [I believe this happened at least twice when we tried mixing stream and non-stream.]
Rich.
-- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com Fedora Windows cross-compiler. Compile Windows programs, test, and build Windows installers. Over 100 libraries supported. http://fedoraproject.org/wiki/MinGW _______________________________________________ epel-devel mailing list -- epel-devel@lists.fedoraproject.org To unsubscribe send an email to epel-devel-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/epel-devel@lists.fedoraproject... Do not reply to spam on the list, report it: https://pagure.io/fedora-infrastructure
On Tue, Mar 1, 2022 at 8:45 AM Stephen John Smoogen smooge@gmail.com wrote:
On Tue, 1 Mar 2022 at 08:19, Richard W.M. Jones rjones@redhat.com wrote:
On Tue, Mar 01, 2022 at 04:21:56AM -0500, Neal Gompa wrote:
On Tue, Mar 1, 2022 at 3:07 AM Richard W.M. Jones rjones@redhat.com wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >= 1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
It's not crazy. EPEL is intended to build on RHEL content, which means we can't depend on something RHEL doesn't publish. If Red Hat wants to publish their buildroot repo, then sure, we could use it.
I wasn't very clear, but I was addressing my remark at Red Hat. There's really no reason why we (Red Hat) don't publish buildroot, in fact my personal view is we ought to for open source reasons.
I do not think you will find much disagreement here.. but after 3+ years of saying it and nothing changing, many of us have made our peace.
To be a bit more fair, we have not blindly added all buildroot content to RHEL. However, we have made progress on coming up with a way to request these packages be added and worked to help teams internally understand the implications of this. We continue to add content to every RHEL minor release.
That's not nothing. It's just not everything.
josh
To keep this a bit more specific, we're trying to build coccinelle for EPEL 9. This requires ocaml [compiler] and a bunch of ocaml packages. They are mainly in RHEL buildroot.
The problem we're going to have (which to be fair is a problem somewhat specific to OCaml linking) is that an alternate OCaml compiler built for EPEL will have different hash values[1] for core libraries.
If we use a different version from the RHEL compiler, say we use Fedora Rawhide version, then all the hashes will be different.
If we use the same version as in RHEL, then most hash values will be the same, but if the compiler flags are even slightly different then there will be some differences.
OCaml libraries in RHEL that we do ship (ocaml-libnbd and others) will not be linkable with code compiled with the EPEL toolchain.
It's entirely possible we don't care about this, and I maybe even agree. But also that people using RHEL with EPEL added will get into weird situations if they try to recompile the virt tools packages.
Rich.
[1] Hash values are computed over the modules to prevent linking incompatible versions:
$ ocamlobjinfo /usr/lib64/ocaml/unix.cma | grep Unix Unit name: Unix 49c6c492a189deeaed5bf77a6793e7fa Unix
and turned into RPM dependencies:
$ rpm -qR ocaml | grep Unix ocaml(Unix) = 49c6c492a189deeaed5bf77a6793e7fa
On Wed, Mar 2, 2022 at 8:52 AM Richard W.M. Jones rjones@redhat.com wrote:
To keep this a bit more specific, we're trying to build coccinelle for EPEL 9. This requires ocaml [compiler] and a bunch of ocaml packages. They are mainly in RHEL buildroot.
The problem we're going to have (which to be fair is a problem somewhat specific to OCaml linking) is that an alternate OCaml compiler built for EPEL will have different hash values[1] for core libraries.
If we use a different version from the RHEL compiler, say we use Fedora Rawhide version, then all the hashes will be different.
If we use the same version as in RHEL, then most hash values will be the same, but if the compiler flags are even slightly different then there will be some differences.
OCaml libraries in RHEL that we do ship (ocaml-libnbd and others) will not be linkable with code compiled with the EPEL toolchain.
It's entirely possible we don't care about this, and I maybe even agree. But also that people using RHEL with EPEL added will get into weird situations if they try to recompile the virt tools packages.
Rich.
[1] Hash values are computed over the modules to prevent linking incompatible versions:
$ ocamlobjinfo /usr/lib64/ocaml/unix.cma | grep Unix Unit name: Unix 49c6c492a189deeaed5bf77a6793e7fa Unix
and turned into RPM dependencies:
$ rpm -qR ocaml | grep Unix ocaml(Unix) = 49c6c492a189deeaed5bf77a6793e7fa
So why not have the OCaml toolchain exposed in RHEL CRB? It sounds like it would be very beneficial to have it there.
On Wed, Mar 02, 2022 at 09:04:04AM -0500, Neal Gompa wrote:
So why not have the OCaml toolchain exposed in RHEL CRB? It sounds like it would be very beneficial to have it there.
It's a good question. I think we chose not to do that simply because we were worried about handling CVEs in a timely way and general support (I'm not sure if CRB is officially supported or not, but there may be some "implicit" support if we're shipping stuff). I would not be opposed to it though.
Rich.
On Thu, Mar 3, 2022 at 12:27 PM Richard W.M. Jones rjones@redhat.com wrote:
On Wed, Mar 02, 2022 at 09:04:04AM -0500, Neal Gompa wrote:
So why not have the OCaml toolchain exposed in RHEL CRB? It sounds like it would be very beneficial to have it there.
It's a good question. I think we chose not to do that simply because we were worried about handling CVEs in a timely way and general support (I'm not sure if CRB is officially supported or not, but there may be some "implicit" support if we're shipping stuff). I would not be opposed to it though.
This requires a Customer Portal account
https://access.redhat.com/solutions/4180391
The CRB repo is documented heavily internally as well.
josh
On Thu, Mar 03, 2022 at 12:42:03PM -0500, Josh Boyer wrote:
On Thu, Mar 3, 2022 at 12:27 PM Richard W.M. Jones rjones@redhat.com wrote:
On Wed, Mar 02, 2022 at 09:04:04AM -0500, Neal Gompa wrote:
So why not have the OCaml toolchain exposed in RHEL CRB? It sounds like it would be very beneficial to have it there.
It's a good question. I think we chose not to do that simply because we were worried about handling CVEs in a timely way and general support (I'm not sure if CRB is officially supported or not, but there may be some "implicit" support if we're shipping stuff). I would not be opposed to it though.
This requires a Customer Portal account
https://access.redhat.com/solutions/4180391
The CRB repo is documented heavily internally as well.
I see "not supported" and "no ABI compatibility guarantee" in bold there which is what we want.
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
Rich.
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
(Insert here my monthly request that there be a single source of truth about packages.)
Rich.
On Fri, Mar 4, 2022 at 6:01 AM Richard W.M. Jones rjones@redhat.com wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
(Insert here my monthly request that there be a single source of truth about packages.)
If you already had a ACKed BZ/Jira for these, you can probably reference that and make the necessary merge requests to comps[1] and pungi[2].
[1]: https://gitlab.com/redhat/centos-stream/release-engineering/comps [2]: https://gitlab.com/redhat/centos-stream/release-engineering/pungi-centos
On Fri, Mar 04, 2022 at 11:01:20AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
https://bugzilla.redhat.com/show_bug.cgi?id=2060850
I didn't know what Product / Component to use, so I used RHEL 9 / distribution.
Rich.
(Insert here my monthly request that there be a single source of truth about packages.)
Rich.
-- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com libguestfs lets you edit virtual machines. Supports shell scripting, bindings from many languages. http://libguestfs.org _______________________________________________ epel-devel mailing list -- epel-devel@lists.fedoraproject.org To unsubscribe send an email to epel-devel-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/epel-devel@lists.fedoraproject... Do not reply to spam on the list, report it: https://pagure.io/fedora-infrastructure
On Fri, Mar 04, 2022 at 11:14:31AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:01:20AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
Currently even with this bug fixed we won't be able to build Coccinelle until RHEL 9.1 is released, which is like 9+ months away. This (if true) is ridiculous. Is there some other solution?
Rich.
V Fri, Mar 11, 2022 at 09:36:08AM +0000, Richard W.M. Jones napsal(a):
On Fri, Mar 04, 2022 at 11:14:31AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:01:20AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
Currently even with this bug fixed we won't be able to build Coccinelle until RHEL 9.1 is released, which is like 9+ months away. This (if true) is ridiculous. Is there some other solution?
That's not ridiculous. That's a release cadence of RHEL.
Maybe you could use EPEL 9 Next which is based on CentOS Stream 9. There is the change already visible http://mirror.stream.centos.org/9-stream/CRB/x86_64/os/Packages/. However, I'm not sure whether EPEL 9 Next is still a thing.
-- Petr
On 11. 03. 22 10:36, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:14:31AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:01:20AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
Currently even with this bug fixed we won't be able to build Coccinelle until RHEL 9.1 is released, which is like 9+ months away. This (if true) is ridiculous. Is there some other solution?
Given that EPEL 9 now builds against CentOS Stream, this is not necessarily true.
The other solution would have been to include the packages in CRB sooner :D
On Fri, Mar 11, 2022 at 10:57:29AM +0100, Miro Hrončok wrote:
On 11. 03. 22 10:36, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:14:31AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:01:20AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
Currently even with this bug fixed we won't be able to build Coccinelle until RHEL 9.1 is released, which is like 9+ months away. This (if true) is ridiculous. Is there some other solution?
Given that EPEL 9 now builds against CentOS Stream, this is not necessarily true.
The other solution would have been to include the packages in CRB sooner :D
I've surely learned a lesson never to use RHEL buildroot for anything.
Rich.
On Fri, Mar 11, 2022 at 09:36:08AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:14:31AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 11:01:20AM +0000, Richard W.M. Jones wrote:
On Fri, Mar 04, 2022 at 10:51:48AM +0000, Richard W.M. Jones wrote:
I'll see if we can move the OCaml packages to CRB. It seems to be the easiest way to fix the original coccinelle build problem.
This gets odder. I see from our internal spreadsheet and downloads that some of the ocaml packages are in fact already in CRB for RHEL 9.0, and others are not. We previously requested that all ocaml-* packages be added to CRB.
Binary packages which are not in CRB but should be:
ocaml-calendar* ocaml-camomile* ocaml-csexp* ocaml-csv* ocaml-curses* ocaml-dune* ocaml-fileutils* ocaml-gettext* ocaml-libvirt* ocaml-source ocaml-xml-light*
Do you need a BZ opened to move these packages to CRB?
Currently even with this bug fixed we won't be able to build Coccinelle until RHEL 9.1 is released, which is like 9+ months away. This (if true) is ridiculous. Is there some other solution?
I have now rebuilt all the packages and they're in a RHEL 9.1 erratum (RHBA-2022:89734-01 for those who have access).
Is there anything we can do to get these packages into the EPEL buildroot so we don't have to wait forever to build coccinelle for EPEL?
Rich.
On Tue, 1 Mar 2022 at 03:06, Richard W.M. Jones rjones@redhat.com wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=2058274
fails to build with:
DEBUG util.py:444: No matching package to install: 'ocaml-dune >= 1.0'
This package is in RHEL 9 buildroot (ocaml-dune-2.8.5-5.el9.x86_64).
I read an earlier thread ("Subject: [EPEL-devel] Re: Packages disappearing from the EPEL 9 buildroot") and it seems to indicate that RHEL 9 buildroot packages aren't going to be available in EPEL 9. This seems crazy, is it really correct?
This is the same as what was done for RHEL-8 and going back a bit to RHEL-5 also since it was not fully published. Buildroot-only packages will need extra care and work to make 'epel-only' versions on them. [Experiments of trying to mix and match CentOS Stream build root and RHEL packages have not gone well in enough cases to keep that up.] EPEL-only packages will be also needed as modules are added to RHEL-9 in various future dot releases.
epel-devel@lists.fedoraproject.org