Re: [MACTCL] Non-Blocking Socket Problem

[Kevan H. <ha...@br...>]

Dear Jeff,

Thanks for looking into the problem.

> I can confirm your problem when running on a Dual-G5, server 8.4.7 (stock) and
> with clients 8.4.7 (all OK), 8.4.14-head and 8.5-head (both hang).

That's a relief.

 > *However*, if I run the server as tclsh (just add a 'vwait forever'
 > after creating the server socket), then I don't get a hang.

I would have tried the same, but I was missing the 'vwait forever' at 
the end. With that line at the end, I was able to duplicate your result.

Server: tclsh 8.4.10 PPC (downloaded from tcltk aqua)
Client: Wish 8.4.13 compiled on Intel 10.4
Result: server fine, client fine.

I ran getblock 4 times, but no freezing (only half-second hesitations).

Server: tclsh 8.4.10 PPC (downloaded from tcltk aqua)
Client: Wish 8.5a4 compiled on Intel 10.4
Result: server fine, client freezes.

So, 8.5a4 does not work with the tclsh.

> In addition, if I run 2
> wishes of the same version level (tested 8.4.14-head), then I also don't hang.

I don't have 8.4.14, so I did what I could with the other versions.

Server: Wish 8.4.13 compiled on Intel 10.4
Client: Wish 8.4.13 compiled on Intel 10.4
Result: server fine, client hangs

At first I thought the client was fine, but on the second run of 
getblock, it froze.

Server: Wish 8.5a4 compiled on Intel 10.4
Client: Wish 8.5a4 compiled on Intel 10.4
Result: server fine, client hangs

Server: Wish 8.12 compiled on G4 10.4
Client: Wish 8.12 compiled on G4 10.4
Result: server fine, client hangs

> Can you please confirm that using the same version on both ends works for you?

It appears that 8.4.14 is unique in not crashing when it works with 
itself. But I can't confirm your result. Do you have a source 
distribution on the net? If not, perhaps you could confirm your result 
once more. It could be that 8.4.14 was lucky with the mouse-clicks.

Yours, Kevan

-- 
Kevan Hashemi, Electrical Engineer
Physics Department, Brandeis University
http://alignment.hep.brandeis.edu/