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: run2.exe segfaults when GDI and X11 elements are present

Ken Brown wrote:
> On 12/28/2009 3:21 AM, Charles Wilson wrote:
>> Fixed in r24. Thanks for the report.
> Yes, that fixed it.


>> I can't reproduce your segfault with r23/r24 and the original dual xml
>> file.
> Did you try test1.xml from
> ?  It's the one where
> mintty is called with the argument '-'.

No, I was using the original test.xml.

The first problem is that test1.xml as posted isn't valid XML -- and the
parser (properly) reports:

run2.exe FATAL: /c/Users/cwilson/test1.xml validation generated an
internal error

but no coredump. Oddly, syntax errors in the attributes of the root
element do not seem to generate meaningful diagnostic messages -- unlike
similar errors elsewhere in the document. This is an libxml2 problem,
and not the "fault" of run2.

OTOH, run2 /should/, however, exit with status 1 rather than status 0 in
this situation. That's a bug -- fixed in r25.

So, first off you need to remove the ';' from the end of the third line
of test1.xml.  Then...

NOW I see the coredump -- which is occurring inside the xml parsing
subroutines. I *HOPE* this is an entirely different bug than the one
that started this thread.

Can you confirm that the original test.xml does NOT coredump anymore,
and that r24 fixes THAT problem?  It's really difficult for me to know
if my fixes are actually FIXING anything, because you keep changing the
test case.  We're now on to the third separate bug [test.xml, test2.xml,
and test1.xml] (four, if you count the 'wrong exit status' above) in the
same thread, with the same subject...

Which is not to say I don't appreciate all the feedback. I'm glad
somebody is really exercising run2 -- BEFORE I lobotomize rxvt and start
"forcing" people to use run2 to replicate current rxvt's 'dual mode
X/GDI' behavior...

>> I'm using Vista...I'll test on XP later today. Can you build with
>> debugging symbols, run under gdb, and generate a backtrace?
> OK, I'll try it later if you can't reproduce the segfault with test1.xml.

Again, does /test.xml/ still coredump for you, or is THAT one fixed:

<?xml version="1.0" encoding="us-ascii"?>
    <Target filename="/usr/bin/mintty.exe" startin="~">
      <Arg>-e /bin/bash</Arg>
    <Target filename="/usr/bin/xterm.exe" startin="~">

I'll followup with another message after I start debugging the new
test1.xml coredump.


Problem reports:
Unsubscribe info:

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