plumber will spin if plumber srv file is removed

plumber will spin if plumber srv file is removed

[Benjamin Purcell]

If the plumber's srv file is removed then the plumber will spin in a
while loop continually reading 0 bytes. The patch at
http://okturing.com/src/3740/body fixes this behavior. I'll apply it
in a few days if there are no objections.

spew

Re: [9front] plumber will spin if plumber srv file is removed

[cinap_lenrek]

will it? if the chan is hung up, you will get an error after
reading past the eof:

		if(q->state & Qclosed){
			if(++q->eof > 3)
				return -1;
			if(*q->err && strcmp(q->err, Ehungup) != 0)
				return -1;
			return 0;
		}

can you reproduce this behaviour?

--
cinap

Re: [9front] plumber will spin if plumber srv file is removed

[Benjamin Purcell]

You're right it does eventually get a -1. I should have checked.

On Tue, Jul 19, 2016 at 12:19 PM,  <cinap_lenrek@felloff.net> wrote:
> will it? if the chan is hung up, you will get an error after
> reading past the eof:
>
>                 if(q->state & Qclosed){
>                         if(++q->eof > 3)
>                                 return -1;
>                         if(*q->err && strcmp(q->err, Ehungup) != 0)
>                                 return -1;
>                         return 0;
>                 }
>
> can you reproduce this behaviour?
>
> --
> cinap

Re: [9front] plumber will spin if plumber srv file is removed

[cinap_lenrek]

the loop is still strange. there should be no reason to try to
read again after getting a eof in theory.

i remember we had issues with apparently getting spurious zero
byte reads, causing cat-v to be fucked a few times a week. tho
that was the only machine encountering these problems... i was
unable to reproduce it. the real cause was never found and
debugging it was difficult.

so we might as well grep all the code for read9pmsg() and make
it consistently thread zero byte reads as eof and give up
immidiately. this is correct thing todo.

--
cinap