Re: Go on sparc

November 12, 2019 10:19 PM, "Stuart Henderson" <[email protected]> wrote:

> On 2019/11/12 11:45, Chris Ross wrote:
> 
>> On Nov 12, 2019, at 00:25, Stuart Henderson <[email protected]> wrote:
>> 
>> sparc64 isn't on upstream's list of supported architectures
>> (https://golang.org/doc/install/source).
>> 
>> There's a third party fork adding this for an old version of go (https://github.com/4ad/go) but
>> even with this it's going to be a lot of work to get something usable, and even then there will be
>> problems building many programs that bundle a "vendored" copy of standard go libraries that will
>> need updating before they can be used.
>> 
>> It's likely to be much easier to rewrite the tools you want in C than it is to get go working.
>> 
>> Well, I don’t know how easy that would be either. These were both, I
> 
> Seriously it's a relatively simple program. Not belittling it, but
> it takes messages via opensmtpd's filter interface, passes them to
> rspamd over http, and feeds the result back to opensmtpd. Porting
> that to a different language is a complete different magnitude of
> complexity than porting a (not-upstream-supported version of) the
> go compiler and standard library to OpenBSD/sparc64.
> 

I very much second that.

The first version of filter-rspamd was very usable, was just a few lines
of code written in a language I didn't (and still don't) know and it was
just a couple hours of work from scratch. If I used a language that I've
known beforehand, it might have taken a few minutes really.

Compare that with porting or maintaining the runtime of a language on an
unsupported architecture. Don't do that :-(


> (...of course it doesn't have to be C, some other language with json+http
> support would also work).
> 

Yes !