[Openmcl-devel] GC and TCP streams, connection errors during writing pending output while GC-finalizing a TCP stream

[Tim Bradshaw]

On 8 Jul 2013, at 09:35, Rainer Joswig wrote:

> But I have no feeling whether the GC should try to write pending stream contents during finalization, while closing the stream

If I wrote something to a stream, but forgot to do a finish-output and then dropped the stream rather than closing it, I think it would probably be polite (least-surprising to badly-written code) of the system to try and write the output before closing the stream.  On the other hand I think it's completely within its rights to decide not to bother on the grounds that I should write code that expects its output to be written properly.  I guess that, if it does want to write then it it should do that wrapped in an error handler which ignores errors (which might be hard in GC?).