Re: T1349 - Fix load order for running with Guile V2 (issue 4849054)

[Ian Hulin]

On 18/08/11 17:06, Patrick McCarty wrote:
> On Thu, Aug 18, 2011 at 3:30 AM, Ian Hulin <address@hidden>
> wrote:
>> On Thu 18 Aug 2011 07:50:28 BST, address@hidden wrote:
>> 
>> The load-order issue appears to be fixed, testing with git and
>> guile 1.8 and 2.0.2. Ignoring whitespace changes, this patch
>> LGTM.
>> 
>> Some more shuffling is needed to make sure we have markup
>> commands defined where they need to be, but that's beyond the
>> scope of this patch.
>> 
>> Should we have a Lilypond Markup Syntax/Guile V2 Tracker, then?
> 
> I think we should have one tracker per issue that we encounter.
> It becomes difficult to handle what has been fixed (and where) when
> we have umbrella issues.
> 
>> If there are any dependencies in the load list that you know
>> about that can still bite us could we record these in the
>> tracker, for instance the markup command itself works with both
>> Guile 1.8 and 2.0 if it is loaded later in the list in a module
>> apart from markup-macros.scm, but fails with 2.0 if you move the
>> definition from markup.scm to the bottom of markup-macros.scm.
> 
> Yes, any issues we find like this should have separate tracker
> issues. I've just opened one:
> 
> https://code.google.com/p/lilypond/issues/detail?id=1826
> 
> I'll open more later today if I find time.
> 
Should we mark Tracker Issue 1055 Verified, and ask Graham for a
Type-Guile-V2-Migration tag in the Tracker system, then?
>> There are also some design issues to consider like:
>> 
>> Would the markup.scm and markup-macros.scm definitions work
>> better as a SCM module?
> 
> Possibly.
> 
Hmm, OK.  Does anyone else have any input on this, Han-Wen, Carl, Jan,
Nicolas, Reinhold (or anyone else with more scheme-fu than me)?

If it's liable to be of real benefit in migrating to Guile V2 then
let's raise a Tracker and I'll add another task to my in-tray.

>> Should we/can we move over to making the markup subsystem
>> hygienic by using define-syntax?
>> 
>> (This is quite a big piece of work, and last time I tried looking
>> at it, it seemed like it should be simple but made my head
>> hurt). It's a translation exercise very like doing the document
>> translation, and currently definitely needs to be done by a human
>> being rather than software.
> 
> If we want backwards compatibility with Guile 1.8, we shouldn't
> move to define-syntax, as it's not implemented in Guile 1.8.  :P
> 

(use-modules (ice-9 syncase))
is available in 1.8.7 and has been incorporated into Guile V2.0
and (sigh) the above call deprecated.

So we *could* do
(if (guile-v2) ( (use-modules (ice-9 syncase))) at the top of the
markup scheme files and then translate them.

>> Current plans once this is pushed is to tackle Tracker 1780
>> (Guile V2 squawking with deprecation errors because of (format)
>> calls without a destination parameter) and then get back to
>> Tracker 1686 (tackling the issue of scheme compilation).
> 
> Sounds good, thanks Ian!

Cheers,
Ian