Re: [MirageOS-devel] [ANN] IPv6 on Mirage!

[Anil Madhavapeddy <anil@xxxxxxxxxx>]

On 20 Nov 2014, at 15:07, Richard Mortier <Richard.Mortier@xxxxxxxxxxxxxxxx> 
wrote:
> 
> 
> On 20 Nov 2014, at 14:52, Thomas Leonard <talex5@xxxxxxxxx> wrote:
>> 
>> It would be useful to know what the problems with using a switch were.
>> It would make packaging the TLS stuff easier if we didn't need e.g.
>> separate gmp.unix/gmp.xen, ctypes.unix/ctypes.xen, etc packages, when
>> the only difference between them is the platform they're compiled for.
>> 
>> Could the mirage tool manage switches for us?
> 
> i wondered about that. i guess it might drive opam appropriately (tracking 
> packages between switches, making sure things updated, etc?) though it might 
> be rather confusing to find that a compilation over here changed the compiler 
> switch (and thus available packages) over there. i've found this with opam 
> packages like merlin for example... (i forget which switch i'm in and 
> suddenly merlin stops working...)

That's another problem indeed.  Since OPAM (and most similar tools, such
as Cabal sandboxes) don't distinguish between host and target tools, we
really need two active switches: one for the host tools like Merlin, annot
and ocp-indent and any code generators (like Atdgen), and another for the
target switch.

> perhaps there's a need to be able to install a package globally (as i believe 
> npm allows for node, at least).

Yeah, this has been brought up before.  It's not quite clear what such a
design would look like in OPAM (a list of ordered switches?) however.

-anil
_______________________________________________
MirageOS-devel mailing list
MirageOS-devel@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel