Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[John Kehayias]

On Tue, Sep 13, 2022 at 01:50 PM, John Kehayias wrote:

>
> I actually added rust-cbindgen-0.23 and 0.24 for another channel which needs them for a
> similar purpose (I will not link the channel but I think you will know which I mean). So I
> can submit the patches here, though I'm not familiar with all the ins and outs of rust
> packaging I can at least take care of the basic updates and needed dependencies.
>

Ah, a bunch of the packaged needed for this (cbindgen dependencies) are done in <https://issues.guix.gnu.org/57702> (unmatched-paren pointed out this patch series on #guix)

So after 57702 I think it might just be one or two packages, then cbindgen and we'll be set for an IceCat update.

> One slight wrinkle is that there is a bug for using version 0.24 for some Firefox versions
> <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259> Though that should be fixed on
> newer ones, I'm not sure if that affects what will end up as IceCat:
> <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259#c5>
>
> Normally I would just go with the latest cbindgen to have that for future uses, but maybe
> we should just have 0.23 and 0.24? I have patches for both, just trying to see what would
> be cleanest here.
>
> Finally, should these be based on staging or master? I hear we'll have staging in master
> in the near future, let me know what will work best for these patches.
>
> John

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[Mark H Weaver]

Hi John,

John Kehayias <john.kehayias@protonmail.com> writes:

> Ah, a bunch of the packaged needed for this (cbindgen dependencies)
> are done in <https://issues.guix.gnu.org/57702> (unmatched-paren
> pointed out this patch series on #guix)
>
> So after 57702 I think it might just be one or two packages, then
> cbindgen and we'll be set for an IceCat update.

That's great news.  Thanks again for your efforts!

      Mark

-- 
Disinformation flourishes because many people care deeply about injustice
but very few check the facts.  Ask me about <https://stallmansupport.org>.

Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[Mark H Weaver]

Hello Guix,

In one week (Sept 20) Mozilla is scheduled to release Firefox 102.3 ESR
with another batch of fixes for security flaws that they will publicize
on the same day, but there will be no corresponding fixes for the 91 ESR
branch.

I hope to update our IceCat package to version 102 by then.

I've identified two dependencies that will be needed, but which are not
yet present on our 'master' branch:

* rust-1.59 (currently on the 'staging' branch)
* rust-cbindgen-0.23

Would anyone be willing to work on adding these?

     Regards,
       Mark

-- 
Disinformation flourishes because many people care deeply about injustice
but very few check the facts.  Ask me about <https://stallmansupport.org>.

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[John Kehayias]

Hello Mark,

On Tue, Sep 13, 2022 at 01:11 AM, Mark H Weaver wrote:

> Hello Guix,
>
> In one week (Sept 20) Mozilla is scheduled to release Firefox 102.3 ESR
> with another batch of fixes for security flaws that they will publicize
> on the same day, but there will be no corresponding fixes for the 91 ESR
> branch.
>
> I hope to update our IceCat package to version 102 by then.
>
> I've identified two dependencies that will be needed, but which are not
> yet present on our 'master' branch:
>
> * rust-1.59 (currently on the 'staging' branch)
> * rust-cbindgen-0.23
>
> Would anyone be willing to work on adding these?
>

I actually added rust-cbindgen-0.23 and 0.24 for another channel which needs them for a similar purpose (I will not link the channel but I think you will know which I mean). So I can submit the patches here, though I'm not familiar with all the ins and outs of rust packaging I can at least take care of the basic updates and needed dependencies.

One slight wrinkle is that there is a bug for using version 0.24 for some Firefox versions <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259> Though that should be fixed on newer ones, I'm not sure if that affects what will end up as IceCat: <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259#c5>

Normally I would just go with the latest cbindgen to have that for future uses, but maybe we should just have 0.23 and 0.24? I have patches for both, just trying to see what would be cleanest here.

Finally, should these be based on staging or master? I hear we'll have staging in master in the near future, let me know what will work best for these patches.

John

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[Maxime Devos]

[-- Attachment #1.1.1: Type: text/plain, Size: 515 bytes --]



On 13-09-2022 15:50, John Kehayias wrote:
> Normally I would just go with the latest cbindgen to have that for future uses, but maybe we should just have 0.23 and 0.24? I have patches for both, just trying to see what would be cleanest here.

 From what I've heard (and experienced in antioxidant), they are 
backwards compatible, so just the latest should be fine, though possibly 
firefox's Cargo.toml (assuming it uses those) needs to be patched to 
accept 0.24 instead of 0.23.

Greetings,
Maxime

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 929 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 236 bytes --]

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[John Kehayias]

Hi Maxime,

On Tue, Sep 13, 2022 at 04:30 PM, Maxime Devos wrote:

>
> On 13-09-2022 15:50, John Kehayias wrote:
>> Normally I would just go with the latest cbindgen to have that for future uses, but
>> maybe we should just have 0.23 and 0.24? I have patches for both, just trying to see
>> what would be cleanest here.
>
> From what I've heard (and experienced in antioxidant), they are backwards compatible, so
> just the latest should be fine, though possibly firefox's Cargo.toml (assuming it uses
> those) needs to be patched to accept 0.24 instead of 0.23.

I think you missed the earlier part where I linked to an upstream bug for building some versions of firefox with cbindgen > 0.23. This is fixed on the latest versions, but I'm not sure for the ESR ones and thus for IceCat. Probably just need to try with newer cbindgen to see what happens. It was not an issue with cargo.toml but an actual build failure.

John

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[Maxime Devos]

[-- Attachment #1.1.1: Type: text/plain, Size: 192 bytes --]



On 13-09-2022 16:35, John Kehayias wrote:
> I think you missed the earlier part where I linked to an upstream bug for building some versions of firefox with cbindgen > 0.23.
Oops, yes.

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 929 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 236 bytes --]

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[Mark H Weaver]

Hi John,

John Kehayias <john.kehayias@protonmail.com> writes:

> I actually added rust-cbindgen-0.23 and 0.24 for another channel which
> needs them for a similar purpose (I will not link the channel but I
> think you will know which I mean). So I can submit the patches here,
> though I'm not familiar with all the ins and outs of rust packaging I
> can at least take care of the basic updates and needed dependencies.
>
> One slight wrinkle is that there is a bug for using version 0.24 for
> some Firefox versions
> <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259> Though that
> should be fixed on newer ones, I'm not sure if that affects what will
> end up as IceCat:
> <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259#c5>
>
> Normally I would just go with the latest cbindgen to have that for
> future uses, but maybe we should just have 0.23 and 0.24? I have
> patches for both, just trying to see what would be cleanest here.

Thanks for pointing this out.  IceCat 102 will almost surely be affected
by the same issues that affect Firefox 102 ESR, so I guess we'll need
rust-cbindgen-0.23 until a better solution is found.

> Finally, should these be based on staging or master? I hear we'll have
> staging in master in the near future, let me know what will work best
> for these patches.

Hmm.  That's a good question.

I guess that the work of properly integrating rust-cbindgen-0.23 into
Guix proper would most naturally be done on the 'staging' branch, since
that branch already has pending changes to our rust packages.

However, there's a problem: we absolutely must have these packages on
the 'master' branch by the morning of September 20th.  I don't know that
we can predict with certainty that 'staging' will be ready to merge by
then.

Therefore, I think we also need to temporarily add these packages to
'master' before the 'staging' merge happens.

I see two possible approaches:

(1) If it's easy to do, and the risk of unintended breakage seems low,
    selected changes could be cherry-picked from 'staging' to 'master',
    taking care to facilitate the upcoming merge.

(2) Alternatively, we could temporarily add private package definitions
    to (gnu packages gnuzilla) on 'master' for the exclusive use of
    IceCat 102, as a stopgap measure.  These could be removed after the
    required packages have been properly integrated into 'master'.

It seems to me that (1) is the right approach for 'rust-1.59', because I
guess we can simply copy the definitions of 'rust-1.58' and 'rust-1.59'
from 'staging' to 'master', while leaving the 'rust' variable unchanged,
and exporting 'rust-1.59'.

I don't know which approach is best for rust-cbindgen-0.23.  It depends
on how much work has been done on the dependencies of rust-cbindgen on
the 'staging' branch.  If in doubt, (2) might be the right approach.

What do you think?

Thanks very much for working on it.

      Regards,
        Mark

-- 
Disinformation flourishes because many people care deeply about injustice
but very few check the facts.  Ask me about <https://stallmansupport.org>.

Re: Needed for IceCat-102: rust-1.59 and rust-cbindgen-0.23

[John Kehayias]

Hi Mark,

Sorry, lost track of the dates on this!

On Tue, Sep 13, 2022 at 04:06 PM, Mark H Weaver wrote:

> Hi John,
>
> John Kehayias <john.kehayias@protonmail.com> writes:
>
>> I actually added rust-cbindgen-0.23 and 0.24 for another channel which
>> needs them for a similar purpose (I will not link the channel but I
>> think you will know which I mean). So I can submit the patches here,
>> though I'm not familiar with all the ins and outs of rust packaging I
>> can at least take care of the basic updates and needed dependencies.
>>
>> One slight wrinkle is that there is a bug for using version 0.24 for
>> some Firefox versions
>> <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259> Though that
>> should be fixed on newer ones, I'm not sure if that affects what will
>> end up as IceCat:
>> <https://bugzilla.mozilla.org/show_bug.cgi?id=1773259#c5>
>>
>> Normally I would just go with the latest cbindgen to have that for
>> future uses, but maybe we should just have 0.23 and 0.24? I have
>> patches for both, just trying to see what would be cleanest here.
>
> Thanks for pointing this out.  IceCat 102 will almost surely be affected
> by the same issues that affect Firefox 102 ESR, so I guess we'll need
> rust-cbindgen-0.23 until a better solution is found.
>
>> Finally, should these be based on staging or master? I hear we'll have
>> staging in master in the near future, let me know what will work best
>> for these patches.
>
> Hmm.  That's a good question.
>
> I guess that the work of properly integrating rust-cbindgen-0.23 into
> Guix proper would most naturally be done on the 'staging' branch, since
> that branch already has pending changes to our rust packages.
>
> However, there's a problem: we absolutely must have these packages on
> the 'master' branch by the morning of September 20th.  I don't know that
> we can predict with certainty that 'staging' will be ready to merge by
> then.
>
> Therefore, I think we also need to temporarily add these packages to
> 'master' before the 'staging' merge happens.
>
> I see two possible approaches:
>
> (1) If it's easy to do, and the risk of unintended breakage seems low,
>     selected changes could be cherry-picked from 'staging' to 'master',
>     taking care to facilitate the upcoming merge.
>
> (2) Alternatively, we could temporarily add private package definitions
>     to (gnu packages gnuzilla) on 'master' for the exclusive use of
>     IceCat 102, as a stopgap measure.  These could be removed after the
>     required packages have been properly integrated into 'master'.
>
> It seems to me that (1) is the right approach for 'rust-1.59', because I
> guess we can simply copy the definitions of 'rust-1.58' and 'rust-1.59'
> from 'staging' to 'master', while leaving the 'rust' variable unchanged,
> and exporting 'rust-1.59'.
>
> I don't know which approach is best for rust-cbindgen-0.23.  It depends
> on how much work has been done on the dependencies of rust-cbindgen on
> the 'staging' branch.  If in doubt, (2) might be the right approach.
>
> What do you think?
>

I see you took the packages that existed for these and put on the gnuzilla-updates branch here. Thanks for doing that, sorry I didn't respond quicker. Hopefully you saw that it was pretty straight forward, a few rust and rust packages updates/new inclusions, but nothing needed anything special.

Does everything build okay for the new icecat now? Do you need anything else?

If I remember, with staging we will only then remove the local rust-1.58 and rust-1.59 packages. The inclusion of the newer rust-cbindgen will also mean a lot of <https://issues.guix.gnu.org/57702> is in gnuzilla.scm and can be coordinated with that patch series. (I'm not on that series but can send a message to that thread.)

> Thanks very much for working on it.
>

Well, thanks for doing the work of getting things moving over here! Hopefully that previous work made it easier. Sorry for dropping the ball earlier.

John