Bug #49850 for Net-SSH-Perl: Re: Win32 support for Net::SSH::Perl

>> Curtis, >> I see you opened a ticket on Net::SSH::Perl recently and stated that you >> intend to include this module in Strawberry in the October release. >> >> I wonder if you have had any success in getting this module to work.  I >> can >> compile and install it, and with a bit of hacking got it to ALMOST work.

>> I have a need for SSH1 and SSH2 to support Router and Switch maintenance, >> and this support needs to be cross platform. >> >> I have not opened a ticket on CPAN, because from what I can see, this >> module >> is no longer supported, which is a shame because there is nothing to >> fully >> replace it.  Net::SSH2 can do SSH2 protocol, but not SSH1.  Anyway, with >> this in mind, I have tried to look at it myself to see if I could get it >> to >> work in Strawberry, because it meets my needs on every other platform.

>> The module out of the box only supports Unix.  i.e. It makes assumptions >> about $HOME and such.  I have played with it, and fixed all of that stuff

>> and have it almost working in Windows, but it seems that it gets no data >> from the socket once it gets to 'client_loop', which is odd since it does >> logon and do authentication, so data is flowing until it gets to >> 'client_loop', and then it seems to loop between 'prepare_channels' and >> 'prepare_for_select' because it seems that $rb and $wb (which are >> pointers >> to anonomous arrays) are always empty.  ..  Or to put it another way... >> It >> gets as far as 'close_read' (if you run it with debug on) and then loops >> presumably because it never empties the buffer.  In an effort to debug >> the >> code I found that 'prepare_for_select' is entered once with ((!$rfd) && >> CHAN_INPUT_OPEN), then on the next loop (($rfd) && CHAN_INPUT_CLOSED), >> but >> since it only calls add() when (($rfd) && CHAN_INPUT_OPEN), add() never >> gets >> called. >> >> There is also Net::SSH::Win32Perl, but that code seems to change nothing >> at >> all.  I guess it gets past the blocking() problem which I solved In the >> native code by simply testing '$^O'  It still loops at 'close_read' >> trying >> to read the data.  Maybe this used to work with some ancient version and >> some change has broken it in the loop above so that it no longer works. >>  Anyway, I decided to ignore it completely and see if I could hack the >> native module into submission, but I am not good enough at debugging PERL >> to >> get farther than I have, so I thought I would pass my findings onto you >> to >> see if my findings would help. >> >> I think it would work OK if the looping problem were fixed.  (Maybe it >> should use IO::Socket::INET instead of the native socket interface to >> easily >> get past the non-blocking problem since I think IO::Socket::INET >> 'normalizes' the blocking interface.  The Native interface requires >> different handling for Windows...  Which is how I fixed it, but then it >> is >> harder to make it platform independent.) >> >> In the end, the simple question is has anyone ever figured out how to get >> Net::SSH::Perl to work in Windows?