From evenson.not.org at gmail.com  Sun Nov  1 15:20:52 2020
From: evenson.not.org at gmail.com (Mark Evenson)
Date: Sun, 1 Nov 2020 16:20:52 +0100
Subject: ABCL uses to recompile LCONS* form
Message-ID: <2810B388-3DC5-4A74-8E25-4C048330A17E@gmail.com>

Note to all:  the abcl.lisp implementation inadvertenly contains a reference to
LCONS*.

<https://github.com/slime/slime/blob/master/swank/abcl.lisp#L1086>

Recompiling the form via SLIME once things have been loaded works fine.

Any chance we can get a standard mechanism to recompile stuff in the
implementation to declare that a given thunk should be recompiled once SWANK
has finished loading?

Back in 2017 or so, @aruttenburg had a patch to some of move the some of the
utlities to load before the implementation-specific, but @stastats objected on
the grounds that we want the initial load to be as fast and error free as
possible.

I really don’t want to redefine LCONS* in the ABCL specific code, but that will
work for me if no one has a better idea.

regards,
Mark

--
"No, this is not a disentanglement, but a progressive /knotting-into/."



From stassats at gmail.com  Sun Nov  1 16:21:26 2020
From: stassats at gmail.com (Stas Boukarev)
Date: Sun, 1 Nov 2020 19:21:26 +0300
Subject: ABCL uses to recompile LCONS* form
In-Reply-To: <2810B388-3DC5-4A74-8E25-4C048330A17E@gmail.com>
References: <2810B388-3DC5-4A74-8E25-4C048330A17E@gmail.com>
Message-ID: <CAF63=13T630CP0Nt7Wc4RaNdKr28HAQprj7NMLWG8bonosyCqA@mail.gmail.com>

On Sun, Nov 1, 2020 at 6:22 PM Mark Evenson <evenson.not.org at gmail.com> wrote:
>
> Note to all:  the abcl.lisp implementation inadvertenly contains a reference to
> LCONS*.
>
> <https://github.com/slime/slime/blob/master/swank/abcl.lisp#L1086>
>
> Recompiling the form via SLIME once things have been loaded works fine.
>
> Any chance we can get a standard mechanism to recompile stuff in the
> implementation to declare that a given thunk should be recompiled once SWANK
> has finished loading?
>
> Back in 2017 or so, @aruttenburg had a patch to some of move the some of the
> utlities to load before the implementation-specific, but @stastats objected on
> the grounds that we want the initial load to be as fast and error free as
> possible.
I don't remember saying that or follow the logic here.


From evenson.not.org at gmail.com  Sun Nov  1 16:36:12 2020
From: evenson.not.org at gmail.com (Mark Evenson)
Date: Sun, 1 Nov 2020 17:36:12 +0100
Subject: ABCL uses to recompile LCONS* form
In-Reply-To: <CAF63=13T630CP0Nt7Wc4RaNdKr28HAQprj7NMLWG8bonosyCqA@mail.gmail.com>
References: <2810B388-3DC5-4A74-8E25-4C048330A17E@gmail.com>
 <CAF63=13T630CP0Nt7Wc4RaNdKr28HAQprj7NMLWG8bonosyCqA@mail.gmail.com>
Message-ID: <BBE96921-DA26-4099-84A9-1D9587BDC443@gmail.com>



> On Nov 1, 2020, at 17:21, Stas Boukarev <stassats at gmail.com> wrote:
> 
> On Sun, Nov 1, 2020 at 6:22 PM Mark Evenson <evenson.not.org at gmail.com> wrote:
>> 
>> Note to all:  the abcl.lisp implementation inadvertenly contains a reference to
>> LCONS*.

[…]

>> Back in 2017 or so, @aruttenburg had a patch to some of move the some of the
>> utlities to load before the implementation-specific, but @stastats objected on
>> the grounds that we want the initial load to be as fast and error free as
>> possible.
> I don't remember saying that or follow the logic here.

My apologies for incorrectly paraphrasing your position.  

Let me try to restate:  the ABCL implementation would like to use some of the
machinery from swank.lisp, such as the [lazy list macroization][1].   The
changes we requested were admittedly jumbled up in whole bunch of other issues,
which we eventually resolved by judicious trimming [2][3].  How could we ensure
that SWANK:LCONS* is present when we load the abcl implementation specific code?

[1]: <https://github.com/slime/slime/blob/master/swank.lisp#L3065> 
[2]: https://github.com/slime/slime/pull/502/commits/1c4681da458470987393ed5a2a4ab555a7d4d30c
[3]: https://github.com/slime/slime/pull/502/commits/1911dcfbe1dbf3445bb41c0483967095cf3c2046

--
"No, this is not a disentanglement, but a progressive /knotting-into/."



From stassats at gmail.com  Sun Nov  1 16:39:54 2020
From: stassats at gmail.com (Stas Boukarev)
Date: Sun, 1 Nov 2020 19:39:54 +0300
Subject: ABCL uses to recompile LCONS* form
In-Reply-To: <BBE96921-DA26-4099-84A9-1D9587BDC443@gmail.com>
References: <2810B388-3DC5-4A74-8E25-4C048330A17E@gmail.com>
 <CAF63=13T630CP0Nt7Wc4RaNdKr28HAQprj7NMLWG8bonosyCqA@mail.gmail.com>
 <BBE96921-DA26-4099-84A9-1D9587BDC443@gmail.com>
Message-ID: <CAF63=11HAxu0K27FOv_DA7Zy4SmwV4f9Bu96tKH_70GnAi=ELw@mail.gmail.com>

I don't see any problem with moving things to whenever they're needed.

On Sun, Nov 1, 2020 at 7:36 PM Mark Evenson <evenson.not.org at gmail.com> wrote:
>
>
>
> > On Nov 1, 2020, at 17:21, Stas Boukarev <stassats at gmail.com> wrote:
> >
> > On Sun, Nov 1, 2020 at 6:22 PM Mark Evenson <evenson.not.org at gmail.com> wrote:
> >>
> >> Note to all:  the abcl.lisp implementation inadvertenly contains a reference to
> >> LCONS*.
>
> […]
>
> >> Back in 2017 or so, @aruttenburg had a patch to some of move the some of the
> >> utlities to load before the implementation-specific, but @stastats objected on
> >> the grounds that we want the initial load to be as fast and error free as
> >> possible.
> > I don't remember saying that or follow the logic here.
>
> My apologies for incorrectly paraphrasing your position.
>
> Let me try to restate:  the ABCL implementation would like to use some of the
> machinery from swank.lisp, such as the [lazy list macroization][1].   The
> changes we requested were admittedly jumbled up in whole bunch of other issues,
> which we eventually resolved by judicious trimming [2][3].  How could we ensure
> that SWANK:LCONS* is present when we load the abcl implementation specific code?
>
> [1]: <https://github.com/slime/slime/blob/master/swank.lisp#L3065>
> [2]: https://github.com/slime/slime/pull/502/commits/1c4681da458470987393ed5a2a4ab555a7d4d30c
> [3]: https://github.com/slime/slime/pull/502/commits/1911dcfbe1dbf3445bb41c0483967095cf3c2046
>
> --
> "No, this is not a disentanglement, but a progressive /knotting-into/."
>


From evenson.not.org at gmail.com  Sun Nov  1 16:48:27 2020
From: evenson.not.org at gmail.com (Mark Evenson)
Date: Sun, 1 Nov 2020 17:48:27 +0100
Subject: ABCL uses to recompile LCONS* form
In-Reply-To: <CAF63=11HAxu0K27FOv_DA7Zy4SmwV4f9Bu96tKH_70GnAi=ELw@mail.gmail.com>
References: <2810B388-3DC5-4A74-8E25-4C048330A17E@gmail.com>
 <CAF63=13T630CP0Nt7Wc4RaNdKr28HAQprj7NMLWG8bonosyCqA@mail.gmail.com>
 <BBE96921-DA26-4099-84A9-1D9587BDC443@gmail.com>
 <CAF63=11HAxu0K27FOv_DA7Zy4SmwV4f9Bu96tKH_70GnAi=ELw@mail.gmail.com>
Message-ID: <8CB2F18C-45E1-4022-AD2F-729F86FCF57E@gmail.com>



> On Nov 1, 2020, at 17:39, Stas Boukarev <stassats at gmail.com> wrote:
> 
> I don't see any problem with moving things to whenever they're needed.

Alright, I’ll try to summon the cycles to make a decent merge request. 

Thanks for the feedback.


--
"No, this is not a disentanglement, but a progressive /knotting-into/."