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: [ANNOUNCEMENT] CALL FOR TESTING: Cygwin 1.7.10


On 12/8/2011 1:52 AM, Christopher Faylor wrote:
On Wed, Dec 07, 2011 at 10:18:40PM -0500, Ken Brown wrote:
On 12/7/2011 6:13 PM, Christopher Faylor wrote:
On Wed, Dec 07, 2011 at 08:03:22AM -0500, Ken Brown wrote:
On 12/6/2011 10:22 PM, Christopher Faylor wrote:
On Tue, Dec 06, 2011 at 09:51:32PM -0500, Ken Brown wrote:
On 12/6/2011 3:02 PM, Corinna Vinschen wrote:
On Dec 6 13:28, Ken Brown wrote:
On 12/6/2011 12:57 PM, Corinna Vinschen wrote:
On Dec 6 12:37, Ken Brown wrote:
On 12/6/2011 4:37 AM, Corinna Vinschen wrote:
A lot of changes and fixes have been made in Cygwin since 1.7.9 has
been released, so we're looking forward to release Cygwin 1.7.10 soon.

Please test the latest developer snapshots at http://cygwin.com/snapshots/
which should have "Release Candidate" quality.

Here's an easily reproducible problem with the 2011-12-05 cygwin1.dll snapshot:

1. Download cygwin-inst-20111205.tar.bz2 .

2. Start mintty from the Cygwin Terminal shortcut.

3. tar -C/ -xf /path/to/cygwin-inst-20111205.tar.bz2
--exclude=usr/bin/cygwin1.dll

An hourglass appears and tar doesn't run.  I'm on W7-64 if that
makes a difference.

Works for me. I just tried from mintty with bash as login shell as well as with tcsh as login shell. I even left out the --exclude and it worked ;)

Does that really occur reproducible on your machine, every try?

Yes. And I just tried twice more with the same behavior. Earlier today I had a similar thing occur on a different machine, on which the 2011-12-04 dll was installed instead of the 2011-12-05 dll. On that machine (again W7-64), the mintty window simply disappeared when I gave the tar command.

Hmm, that's weird.


The only other detail I can think of is that I started mintty by
right-clicking on the shortcut and choosing "Run as administrator".

Still works for me. That's puzzeling. Any chance to debug that? Stracing, attaching with GDB, building your own debug-enabled Cygwin, etc?

I ran the tar command under strace. I'm not good at deciphering strace output, but you can see it here:

http://www.math.cornell.edu/~kbrown/cygwin/strace.out

A few further observations:

- As the strace output shows, tar does in fact run, contrary to what I
originally thought; it's just that there's no indication in the terminal
that tar finished.

- When I finally get tired of waiting and kill mintty (and the
associated bash process) from another terminal, mintty leaves the
following stackdump on the desktop:

Exception: STATUS_ACCESS_VIOLATION at eip=610CD8B1
eax=0000010C ebx=0028C880 ecx=00000000 edx=0000010C esi=00000000
edi=0028C860
ebp=0028C7C8 esp=0028C790 program=C:\cygwin\bin\mintty.exe, pid 4972,
thread main
cs=0023 ds=002B es=002B fs=0053 gs=002B ss=002B
Stack trace:
Frame     Function  Args
0028C7C8  610CD8B1  (0000010C, 0028C95C, 0028C880, 0028C860)
0028C938  610CDC42  (00000002, 0028C95C, 00000000, 00000000)
20052168  610D1FE5  (00630072, 00000000, 00000018, 00000033)
200524F0  00000000  (6E776F72, 6D64612D, 2E2F6E69, 746E696D)
End of stack trace

- Other mintty windows that are open become unusable.  The title bar
shows "not responding".

Does any of this help? Should I be stracing mintty rather than tar?

Apparently yes. I can't reproduce this either. The strace you provided does look perfectly normal.

OK. I started mintty from the Cygwin Terminal shortcut, attached strace to it in a second terminal, and ran the tar command in the first terminal. The strace output is at

http://www.math.cornell.edu/~kbrown/cygwin/strace_mintty.out

Here's one more data point: The problem occurs only in a mintty started
>from the shortcut. If I start mintty from an existing mintty, I can run
the tar command without a problem.

Still can't duplicate it.


I can see that mintty is having problems but it isn't clear why.

I'm going to generate a new snapshot with more debugging.  Could you
give it a try when it shows up?

I'd like to see both the above stack dump, if you get it and another
snapshot.

Here's the stack dump:


Exception: STATUS_ACCESS_VIOLATION at eip=610CD8B1
eax=0000010C ebx=0028C880 ecx=00000000 edx=0000010C esi=00000000
edi=0028C860
ebp=0028C7C8 esp=0028C7A0 program=C:\cygwin\bin\mintty.exe, pid 4396,
thread main
cs=0023 ds=002B es=002B fs=0053 gs=002B ss=002B
Stack trace:
Frame     Function  Args
0028C7C8  610CD8B1  (0000010C, 0028C95C, 0028C880, 0028C860)
0028C938  610CDD87  (00000002, 0028C95C, 00000000, 0028C860)
20052208  610D2045  (00630072, 00000000, 00000018, 00000033)
20052590  00000000  (6E776F72, 6D64612D, 2E2F6E69, 746E696D)
End of stack trace

And the strace output is at

http://www.math.cornell.edu/~kbrown/cygwin/strace_20111208_snap.out

By the way, ssh doesn't work with this snapshot; I get the error message
"PRNG is not seeded" when I run "ssh user@host", and I also can't start
sshd.  (/var/log/sshd.log contains the same "PRNG is not seeded" error
message.)

Sigh. I screwed up select() so the test is suspect.

No problem. I redid the test:


$ uname -a
CYGWIN_NT-6.1-WOW64 fiona 1.7.10s(0.255/5/3) 20111208 06:50:31 i686 Cygwin

$ cat ~kbrown-admin/*.stackdump
Exception: STATUS_ACCESS_VIOLATION at eip=610CD801
eax=0000010C ebx=0028C880 ecx=00000000 edx=0028C95C esi=00000000 edi=0028C860
ebp=0028C7C8 esp=0028C7A0 program=C:\cygwin\bin\mintty.exe, pid 6532, thread main
cs=0023 ds=002B es=002B fs=0053 gs=002B ss=002B
Stack trace:
Frame Function Args
0028C7C8 610CD801 (0000010C, 0028C95C, 0028C880, 0028C860)
0028C938 610CDB76 (00000002, 0028C95C, 00000000, 00000000)
20052208 610D1F25 (00630072, 00000000, 00000018, 00000033)
20052590 00000000 (6E776F72, 6D64612D, 2E2F6E69, 746E696D)
End of stack trace


The strace output is still at

http://www.math.cornell.edu/~kbrown/cygwin/strace_20111208_snap.out


-- 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]