This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: cygwin on W7: stalled scp (openssh-6.8p1), lost ssh-scp-pipe data




On 6/12/2015 2:59 PM, Theodor.Kazhan@gmx.de wrote:
Hi Corinna, all,

in case someone is still as curious about it as I am, it took nearly one week and 3433 attempts:

$  while true; do DATE=`date +%Y%m%d_%H%M%S`; echo -n "started: $DATE"; scp -v -C -o BatchMode=yes tester@10.IP2.IP3.IP4:"logDownload_m/TEST.txt" . > scp_$DATE.txt 2>&1; echo " - finished: `date +%Y%m%d_%H%M%S`"; rm -v `ls -1 scp_2015* | head -n -3`; sleep 5; done
started: 20150605_170918 - finished: 20150605_171220
âscp_20150603_182927.txtâ entfernt
started: 20150605_171225 - finished: 20150605_171531
âscp_20150603_183041.txtâ entfernt
started: 20150605_171536 - finished: 20150605_171842
âscp_20150603_183156.txtâ entfernt
. . .
started: 20150612_113735 - finished: 20150612_113845
âscp_20150612_113347.txtâ entfernt
started: 20150612_113850 - finished: 20150612_114003
âscp_20150612_113502.txtâ entfernt
started: 20150612_114009 <== this ssh-download is stalled

... to reproduce this issue one more time on my 2nd env (cygcheck attached to my previous email dated 05. June 2016, 17:00). This time, a 32k buffer (the 16k-blocks B003363 and B003364) is lost:

$ sdiff -stw40 <(cat TEST.txt | sed -re "s/(B0.....-N001024-)/\1\n/g" | cut -d- -f-2) <(seq -w 1 1 94796 | sed -re "s/.*/B0&-N000001/g")
.                   >  B003363-N000001
.                   >  B003364-N000001
.B094793-N000001    <
.B094794-N000001    <


Any advice is still highly appreciated, e.g. on howto debug into Cygwin/Win I/O-APIs, read/write functions, pipe implementation... For much more details, please refer to my previous emails.


Thanks and BR,
T.


Hi Theodor,
Question: last cygwin 2.0.4  or previous version ?

Maybe is no relevant for your case but there were some fixes

In theory during tests a 64K buffer should be faster

Regards
Marco



--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]