www.libssh2.org | Daily snapshots | Mailing list archive | Docs | Examples | github

Archive Index This month's Index

Subject: Re: libssh2_channel_read issues

Re: libssh2_channel_read issues

From: Zulfiqar Malik <zulfiqar.malik_at_gmail.com>
Date: Thu, 12 Apr 2012 15:00:00 +0100

I am actually using libssh2_channel_write. I am basically just doing a cd
<somedir> and then reading the output. The cd finishes and yet I am stuck
in my loop since I never receive an EOF.
I can use waitsocket. The Sleep is just test code. waitsocket doesn't
imrpove the situation.

On Thu, Apr 12, 2012 at 11:25 AM, Matthew Booth <mbooth_at_redhat.com> wrote:

> On 04/11/2012 05:55 PM, Zulfiqar Malik wrote:
>
>> I am trying to write a very small shell client for some automation
>> related things and am running into problems with libssh2_channel_read.
>> Here are the two scenarios (application has to be single threaded.):
>> 1. Blocking mode: libssh2_channel_read blocks if there's no data and
>> never returns. Its impossible to use this function reliably.
>> 2. Non-blocking mode: libssh2_channel_read doesn't block but keeps on
>> returning -37 (LIBSSH2_ERROR_EAGAIN) and I end up with essentially the
>> same problem i.e. stuck in an infinite loop. To top it off
>> libssh2_channel_eof doesn't work and always returns 0.
>>
>
> You don't say what's happening at the other end. Can we assuming it's
> something like:
>
> libssh2_channel_exec(channel, "cat foo");
>
> and that you want to exit when the contents of foo have been returned? Is
> is possible, for example, that your channel is not returning eof because it
> isn't eof?
>
>
> I don't know what's wrong or what needs to be done but I cannot seem to
>> find a way through this problem. I have tried libssh2 v 1.3.1 and the
>> latest one 1.4.1 with the same result. Given below is my readOutput
>> method which reads the output:
>> for (;;count++)
>> {
>> /* loop until we block */
>> int rc;
>> do {
>> char buffer[1024];
>> rc = libssh2_channel_read(channel, buffer, sizeof(buffer));
>> if (rc > 0) {
>> bytecount += rc;
>> buffer[rc] = '\0';
>>
>
> Probably not related, but ^^^ corrupts the stack if libssh2_channel_read
> returns a full buffer. If you're going to explicitly null terminate the
> buffer, you need:
>
> rc = libssh2_channel_read(channel, buffer, sizeof(buffer) - 1);
>
>
> fprintf(stderr, "%s", buffer);
>> count = 0;
>> }
>> else {
>> if (rc != LIBSSH2_ERROR_EAGAIN)
>> fprintf(stderr, "libssh2_channel_read returned
>> %d\n", rc);
>> }
>> if (libssh2_channel_eof(channel)) {
>> rc = 0;
>> break;
>> }
>> }
>> while(rc > 0);
>> //::Sleep(3);
>> /* this is due to blocking that would occur otherwise so we loop
>> on
>> this condition */
>> if(rc == LIBSSH2_ERROR_EAGAIN)
>> {
>> ::Sleep(5);
>> //waitsocket(sock, sshSession);
>>
>
> I'm assuming this references something like the waitsocket() example at
> http://www.libssh2.org/**examples/ssh2_exec.html<http://www.libssh2.org/examples/ssh2_exec.html>.
> Why the sleep instead?
>
> }
>> else
>> break;
>> }
>>
>
> Matt
> --
> Matthew Booth, RHCA, RHCSS
> Red Hat Engineering, Virtualisation Team
>
> GPG ID: D33C3490
> GPG FPR: 3733 612D 2D05 5458 8A8A 1600 3441 EA19 D33C 3490
>

-- 
Regards,
Zulfiqar Inayat Malik.

_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2012-04-12

the libssh2 team