[m-rev.] for possible review: use explicit streams in the MLDS backend

[Julien Fischer]

On Tue, 17 Nov 2020, Zoltan Somogyi wrote:

> 2020-11-16 23:40 GMT+11:00 "Julien Fischer" <jfischer at opturion.com>:
>>> It has bootchecked in hlc.gc, csharp and java grades. In the last two
>>> grades, it had a few more test failures than the last bootchecks I did
>>> in those grades, but I haven't seen any evidence that they were caused
>>> by this change.
>>
>> If it was the string format tests, I committed a fix for them yesterday.
>
> No, it wasn't. The diff between the lists of failed tests from several days
> ago and with the diff is attached.

Are these being run at the standard optimisation level?

> However, I have now done another bootcheck in java grade in another
> workspace, and the only differences in the stage 2 .java files between
> that bootcheck and the bootcheck of this diff are in the modules whose
> source code this diff changes, so I am pretty sure this diff did not cause
> the extra failures.

I've just run a comparison between rotd-2020-11-16 and the current
repository head for the Java grades and do not get any additional
failures.  (The only tests that are failing are ones that I would
expect to fail, due to things like the RTTI implementation being
incomplete.)

Julien.