[Verse-dev] Verse R4pre

[Thorsten Jordan]

Eskil Steenberg wrote:
> Hi
> 
> I have downloaded your apps and the binarys crash tor me too. I havent 
> compiled and tested them my self yet. It would take me some time to 
> compile and debug the code, could you send me some outputs and show me 
> in what line the code crashes?
i've uploaded gzip'ed output to
www.student.informatik.tu-darmstadt.de/~thjordan/vs.txt.gz
and
www.student.informatik.tu-darmstadt.de/~thjordan/push.txt.gz
it is the output of the server and my pusher. You will see that the 
pusher creates links but the server never receives them. Later my viewer 
is connected to the server, but has nothing to view, because there are 
no links.

I tried to find out why the server crashed sometimes (it happened on at 
least two computers), but this time it didn't crash, although i'm sure i 
redid the same things like when it crashed. But i've encountered another 
thing: when pushing the bunny or the cow the pusher once crashed in 
v_pack.c, line 24. Maybe that helps a bit.

I'll be on vacation from sunday for 7-10 weeks, so i can't give any more 
precise information until then. I hope anyone will try my apps and 
report some experiences or even bugs with them...

> Port bugs are strange, In my experience the most usual bugs that appear 
> when porting is un initialized values (Linux often clears memory) and 
> stack overflow bugs. (That was the kind of bug Brecht found). But as it 
> sounds on your description of the problems to be a  an issue if 
> performance where the sending just stops after a while and you dont get 
> all your data  sent. So its not a particular place or function that is 
> broken. This problem is the main reason for the large re-write in R4.
uninitialized values or stack overflow bugs can also be the reason for 
this, especially with the crash in v_pack.c, line 24. in that line some 
memory is changed using pointers and pointer arithmetic, which is highly 
dangerous and often the cause of bugs... that's why i use C++ instead of 
C and no Arrays. Vectors are much safer.

> 
> So my question now is would you be interested in porting your app to R4 
> and see if it works in Linux and works better under Win32?
yes, if R4 seems to be at least as stable as R3P2, but sooner or later 
sure, i need to be up to date with the protocol.
I also downloaded and unpacked the r4pre-zip, but there wasn't at least 
a makefile in it, so i didn't looked at it any further. Is the spec 
updated for R4? is there an API-changes list?

-- 
Greetings, Thorsten