Call for testing: OpenSSH-6.5

[Darren Tucker]

On Sat, Jan 25, 2014 at 3:52 PM, Hisashi T Fujinaka <htodd at twofifty.com> wrote:
> On Sat, 25 Jan 2014, Darren Tucker wrote:
>
>> Maybe there's something wrapping write() on your system too?  Try adding
>> this to the start of atomicio.c:atomicio6():
>>
>> printf(stderr, "read %x write %x argument %x\n", read, vwrite, f);
>>
>> then compile it and run a small scp:
>>
>> $ ./scp testfile localhost:/tmp/
>> read b73eced0 write b73ecf50 argument b73eced0
>>
>> The patch below makes the first argument to atomicio an int rather than
>> a function pointer.
>>
>> On the plus side: it also allows the removal of some hacks around
>> vwrite/writev.  On the minus side: it's a large change albeit mostly
>> mechanical.
>
>
> I tried building after applying the patch, but I think I started in the
> wrong place? I think I started with a fresh git clone.

I think you started with a tree which had one of the changes I
suggested upthread ( where I suggest changing the line to:

pfd.events = f == vwrite ? POLLOUT : POLLIN;

> Also, the scp told me it couldn't find /usr/local/bin/ssh. I'm guessing
> there are environment variables I need.

you can work around that with "./configure --prefix=/usr" for testing,
but don't do "make install" unless you really want to replace the
system binaries in /usr/{bin,sbin}

> atomicio.c:59:57: error: 'vwrite' undeclared (first use in this function)

"vwrite" doesn't appear in the unmodified version of atomicio.c except
in a comment.

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
    Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.