[9fans] APE's main

[9fans] APE's main

[Federico G. Benavento]

hola,

I've been porting pdcurses using draw(2) and event(2),
the thing is that einit() creates some procs that are killed when
exit() is called, and I think APE's main wasn't doing it.
since I'm an ignorant on the subject at hand I brute forced
/sys/src/ape/lib/ap/386/main9.s and now it seems to work
as spected. My question is:  "does this look right" ?

lotte% yesterday -d /sys/src/ape/lib/ap/386/main9.s
diff /n/dump/2007/0426/sys/src/ape/lib/ap/386/main9.s /sys/src/ape/lib/ap/386/main9.s
10a11
> 	MOVL	$_exits<>(SB), AX
13a15,17
> 
> DATA	_exits<>+0(SB)/4, $"main"
> GLOBL	_exits<>+0(SB), $5

thanks

Federico G. Benavento

---
/bin/fortune:
*** {02.38.001} Win - Multiple Trillian vulnerabilities

note: I'm porting lynx as a test case for the curses port
if anyone wants to play with it a binary and some cfg files
that should be in /tmp are at:
http://plan9.homeip.net/magic/webls?dir=/misc

Re: [9fans] APE's main

[Russ Cox]

> lotte% yesterday -d /sys/src/ape/lib/ap/386/main9.s
> diff /n/dump/2007/0426/sys/src/ape/lib/ap/386/main9.s /sys/src/ape/lib/ap/386/main9.s
> 10a11
>> 	MOVL	$_exits<>(SB), AX
> 13a15,17
>> 
>> DATA	_exits<>+0(SB)/4, $"main"
>> GLOBL	_exits<>+0(SB), $5

> My question is:  "does this look right" ?

no, not at all.  you are passing "main" to
the exit function, which expects an integer.
the old one was correct: it passes the return
value of int main() to exit.  

i'm confused about why changing the
exit value would make things start working
anyway.  maybe it didn't really.

my guess would be your actual problem
is that draw and event use atexit, which
ape doesn't provide.  i'm surprised that
you don't get link errors.

russ

Re: [9fans] APE's main

[Federico G. Benavento]

> no, not at all.  you are passing "main" to
> the exit function, which expects an integer.
> the old one was correct: it passes the return
> value of int main() to exit.  
> 
yeah, now I see it

> i'm confused about why changing the
> exit value would make things start working
> anyway.  maybe it didn't really.
> 
I have no idea, what I know is that if I manually do
exit(0) in main() the event procs are killed by atexit()
and without exit(0) stay there.

> my guess would be your actual problem
> is that draw and event use atexit, which
> ape doesn't provide.  i'm surprised that
> you don't get link errors.
> 
ape has atexit() /sys/src/ape/lib/ap/stdio/atexit.c 

thanks for the explanation

Federico G. Benavento

---
/bin/fortune:
Please wait. This can take some time.