Re: port-for-each vs lazy sweep

[Ludovic Courtès]

Hi Han-Wen,

Han-Wen Nienhuys <address@hidden> writes:

> Applied, with corrections.

I noticed the (supposedly related) regression compared to 1.8.  Consider
this program:

  (define p (open-output-file "TEST-FILE"))

  (setvbuf p _IOFBF 16384)

  (write "hello" p)

When running it with "guile-1.8 the-program.scm", `TEST-FILE' contains
the string "hello" upon completion.  However, with HEAD, `TEST-FILE' is
empty.

Can you try to see if it somehow relates to your patch?

(I suspect that ports used to be flushed when reclaimed, which is no
longer the case with the weak hash table.)

Thanks in advance,
Ludovic.