Status of Meson support, where are we? (was Re: Build systems and `uinstall` option)
Hi all, it feels like this deserves its own thread. I was a little surprised by the news that we are already ready to deprecate Make. Hopefully, I'm just not up to date and this is the case! If I'm horribly off track, please ignore the following. IIRC, we said this would be the case if we can do everything with Meson what we can do with Make (due to lack of knowing what the latter can, we made a list of target environments that we need to support at least) and Meson produces the same binaries as Make. Are we there yet? My last state wrt. the more special target environments like lib- payload was that it seems feasible, but I haven't seen a PoC yet. On 10.11.22 17:43, Felix Singer wrote:
The Make build system is going to be deprecated with the next release and it is going to be dropped with the release after that.
"next release" being the one that just started or the next after that? iow. does this mean deprecate with flashrom 1.3 and drop in 1.4?
We discussed the Make build system in many meetings now and we all agreed that we won't add any more features to it. That means its current feature state is frozen and it doesn't get any additional functionality, just bug fixes, refactorings or adjustments as far as they are really necessary.
Was this ever discussed on / announced to the ML? My status was that this was of the "nobody objected during some meeting" decision kind. Nico
from the peanut gallery of packagers: first, please issue an actual release - with 100% working meson support including every feature the previous system has (cross) - full docs on how to use it ask packagers to update. given them 3 months at least. fix any problems that arise, including "i used to be able to do X by calling with Y" leading to adding a way to do X. After fixing them release a new micro with the fix. only after everybody has been able to update, is it ok to drop the previous support from the master/main/whatever branch, leading to the next release not having it. If it takes longer until the new system is good enough, that's ok. It is not reasonable to set a hard timetable because that leads to dropping the old system before the new one meets requirements. I have generally (not flashrom) seen a pattern of we don't like automake shiny system X is better add X, release it it works in the linux/x86_64 do-normal-things happy path there are issues with other systems, cross or something the previous system is dropped anyway It's not always like this, but its often enough to make comments like this email!
Hi Nico, On Fri, 2022-11-11 at 14:49 +0100, Nico Huber wrote:
Hi all,
it feels like this deserves its own thread. I was a little surprised by the news that we are already ready to deprecate Make. Hopefully, I'm just not up to date and this is the case! If I'm horribly off track, please ignore the following.
IIRC, we said this would be the case if we can do everything with Meson what we can do with Make (due to lack of knowing what the latter can, we made a list of target environments that we need to support at least) and Meson produces the same binaries as Make.
Are we there yet?
My last state wrt. the more special target environments like lib- payload was that it seems feasible, but I haven't seen a PoC yet.
I didn't say that we are "ready" to deprecate it, which would mean that there is nothing left to do at this moment. There is still work left. What I basically said was that this is our goal for the upcoming 1.3 release.
On 10.11.22 17:43, Felix Singer wrote:
The Make build system is going to be deprecated with the next release and it is going to be dropped with the release after that.
"next release" being the one that just started or the next after that? iow. does this mean deprecate with flashrom 1.3 and drop in 1.4?
Deprecation is planned for 1.3 release. Dropping support for the release after that, so e.g. 1.4 or 2.0.
We discussed the Make build system in many meetings now and we all agreed that we won't add any more features to it. That means its current feature state is frozen and it doesn't get any additional functionality, just bug fixes, refactorings or adjustments as far as they are really necessary.
Was this ever discussed on / announced to the ML? My status was that this was of the "nobody objected during some meeting" decision kind.
As far as I know, this wasn't discussed on the mailing list. Though, this was discussed a lot on the meetings and not just "nobody objected during some meeting". // Felix
Hi Felix, On 11.11.22 17:47, Felix Singer wrote:
On Fri, 2022-11-11 at 14:49 +0100, Nico Huber wrote:
it feels like this deserves its own thread. I was a little surprised by the news that we are already ready to deprecate Make. Hopefully, I'm just not up to date and this is the case! If I'm horribly off track, please ignore the following.
IIRC, we said this would be the case if we can do everything with Meson what we can do with Make (due to lack of knowing what the latter can, we made a list of target environments that we need to support at least) and Meson produces the same binaries as Make.
Are we there yet?
My last state wrt. the more special target environments like lib- payload was that it seems feasible, but I haven't seen a PoC yet.
I didn't say that we are "ready" to deprecate it, which would mean that there is nothing left to do at this moment. There is still work left. What I basically said was that this is our goal for the upcoming 1.3 release.
so there will be further development (not just fixes) on the 1.3.x branch? Honest question because I wasn't present when the decision was made, and the announcement didn't seem to suggest this. Nico
On Fri, 2022-11-11 at 18:52 +0100, Nico Huber wrote:
On 11.11.22 17:47, Felix Singer wrote:
On Fri, 2022-11-11 at 14:49 +0100, Nico Huber wrote:
it feels like this deserves its own thread. I was a little surprised by the news that we are already ready to deprecate Make. Hopefully, I'm just not up to date and this is the case! If I'm horribly off track, please ignore the following.
IIRC, we said this would be the case if we can do everything with Meson what we can do with Make (due to lack of knowing what the latter can, we made a list of target environments that we need to support at least) and Meson produces the same binaries as Make.
Are we there yet?
My last state wrt. the more special target environments like lib- payload was that it seems feasible, but I haven't seen a PoC yet.
I didn't say that we are "ready" to deprecate it, which would mean that there is nothing left to do at this moment. There is still work left. What I basically said was that this is our goal for the upcoming 1.3 release.
so there will be further development (not just fixes) on the 1.3.x branch? Honest question because I wasn't present when the decision was made, and the announcement didn't seem to suggest this.
I just thought about that meeting again and you are right. In the meeting, it was decided that everything what doesn't work will be just reverted if there is no fix available. So there shouldn't be further development other than bug fixes and reverts. Somehow I mixed that up with "We will do a release candidate, see what works and continue from there". Sorry for the mess and confusion. I still have the plans from the last 6 months in my head :) // Felix
Hi, Differences to the Makefile are: * MacOS: No support the internal programmer (the system APIs under DirectHW are deprecated) Open points are: * ni845x on Windows (proprietary dependency) * crosscompiling DOS (djgpp has problems with format-strings; %lu vs. %u) * crosscompiling libpayload (I'm not even able to compile it with make; 1.2 & master) I don't know the plan for the 1.3 relase, if we bring the missing targets in or if it's too late. -- Thomas On Fri, 2022-11-11 at 14:49 +0100, Nico Huber wrote:
Hi all,
it feels like this deserves its own thread. I was a little surprised by the news that we are already ready to deprecate Make. Hopefully, I'm just not up to date and this is the case! If I'm horribly off track, please ignore the following.
IIRC, we said this would be the case if we can do everything with Meson what we can do with Make (due to lack of knowing what the latter can, we made a list of target environments that we need to support at least) and Meson produces the same binaries as Make.
Are we there yet?
My last state wrt. the more special target environments like lib- payload was that it seems feasible, but I haven't seen a PoC yet.
On 10.11.22 17:43, Felix Singer wrote:
The Make build system is going to be deprecated with the next release and it is going to be dropped with the release after that.
"next release" being the one that just started or the next after that? iow. does this mean deprecate with flashrom 1.3 and drop in 1.4?
We discussed the Make build system in many meetings now and we all agreed that we won't add any more features to it. That means its current feature state is frozen and it doesn't get any additional functionality, just bug fixes, refactorings or adjustments as far as they are really necessary.
Was this ever discussed on / announced to the ML? My status was that this was of the "nobody objected during some meeting" decision kind.
Nico _______________________________________________ flashrom mailing list -- flashrom@flashrom.org To unsubscribe send an email to flashrom-leave@flashrom.org
Hello, The plan is (and it was like this from the very beginning :) ) Release 1.3 introduces meson officially, specifically with documentation and description in README We noticed that although meson has been added since 1.2, and has users already, it is documented less than it should be. At this point, there could still be some open points as Thomas mentioned. Release 1.3+1 has meson covering all environments and features that we want. This release will have BOTH meson and makefile, and makefile announced to be deprecated in the next release. Release 1.3+2 drops makefile. Greg thank you for your perspective, very much appreciated! There definitely will be enough time between releases, to allow everyone to adjust. There are more details in the public doc of One Build System Working Group https://docs.google.com/document/d/1TVso_VbrLEbGc7BNrYxqlrP8jMxB23TuMqEI3PnC... -- Anastasia.
Hi Anastasia, On Mon, 2022-11-14 at 21:17 +1100, Anastasia Klimchuk wrote:
The plan is (and it was like this from the very beginning :) )
I feel like this needs correction. It was definitelly the plan from the very beginning. It was discussed multiple times. It's not the plan anymore since you all decided to torpedo the plans about the 1.3 release from the last 6 months in our last meeting ;) But good to see that there is a new plan! // Felix
participants (5)
-
Anastasia Klimchuk -
Felix Singer -
Greg Troxel -
Nico Huber -
Thomas Heijligen