[9fans] suicidal kfs, again

[9fans] suicidal kfs, again

[kazumi iwane]

Hello 9fans,

About a month ago, I reported two problems,
assertion failure in libmemdraw and suicidal kfs.
The libmemdraw problem went away, after I
re-installed from the March 28th release. Thank
you, bell labs people.

Kfs is a lot more stable now, too. Still, it seems
that it is weak under the load. Compiling large
programs like /sys/src/cmd/^(gs tex) kills it.
Another example,
	{while() fortune} > /dev/null
kills kfs in about 5 minutes on my system. Does
it on yours?

I know kfs isn't meant to be the real thing, but
until I get another machine and IDE support in fs
code, I have to live with it. I want to dive into
this problem, but when kfs dies, there seems to
be nothing left I can do afterwards. I am looking for
any debugging strategies and techniques.

Thanks,

- kazumi

Re: [9fans] suicidal kfs, again

[Russ Cox]

I ran
	while(){ mk; mk clean }
in my kfs /sys/src/cmd/gs directory
for a couple hours this morning and
didn't have any problems.  Then again,
I have enough memory that I don't swap
while building gs.  Do you?

Do you know if you're swapping when
running the
	{while() fortune} >/dev/null
?

The fact that kfs dies there is more
disturbing: that should be random-bound
rather than kfs- or memory-bound.

Russ