graph (plotutils) seg-faulting

[Brian Inglis]

On 2020-07-23 11:00, Tony Richardson via Cygwin wrote:
> On Thu, Jul 23, 2020 at 11:49 AM Ken Brown via Cygwin wrote:
>> On 7/23/2020 12:48 AM, Marco Atzeri via Cygwin wrote:
>>> On 23.07.2020 00:12, Ken Brown via Cygwin wrote:
>>>> On 7/22/2020 6:07 PM, Tony Richardson via Cygwin wrote:
>>>>> On Wed, Jul 22, 2020 at 3:48 PM Marco Atzeri via Cygwin wrote:
>>>>>> On 22.07.2020 14:06, Ken Brown via Cygwin wrote:
>>>>>>> On 7/22/2020 1:36 AM, Marco Atzeri via Cygwin wrote:
>>>>>>>> On 22.07.2020 02:10, Tony Richardson via Cygwin wrote:

>>>>>>>>> I have attached both the cygcheck and strace output.

>>>>>>>> something is clearly interfering with your Cygwin installation
>>>>>>>>
>>>>>>>> --- Process 11136 loaded C:\Applications\cygwin64\bin\cygwin1.dll at
>>>>>>>> 0000000000c80000
>>>>>>>> --- Process 11136 unloaded DLL at 0000000000c80000

>>>>>>>
>>>>>>> I can reproduce the problem on my system:
>>>>>>>
>>>>>>> $ strace graph
>>>>>>> ...
>>>>>>> Segmentation fault
>>>>>>>
>>>>>>> I thought that simply rebuilding plotutils might fix the problem,
>>>>>>> but there were a lot of compilation warnings, and the build
>>>>>>> failed. Some of the warnings look serious to me, but I have no
>>>>>>> idea if they could cause Windows to try to load DLLs at strange
>>>>>>> addresses.

>>>>>> I rebuilt and uploaded a test version 2.6-6 only for 64bit
>>>>>> try to see if something change

>>>>> I tried the test version, unfortunately it still segfaults for me.

>>> no surprise. It was unlikely to work

>>>> Same for me.  I'm also getting strange output from ldd, but maybe 
>>>> that's just another manifestation of the same problem:
>>>>
>>>> $ ldd /usr/bin/graph
>>>> ntdll.dll => /c/WINDOWS/SYSTEM32/ntdll.dll (0x7ff82a4c0000)
>>>> KERNEL32.DLL => /c/WINDOWS/System32/KERNEL32.DLL (0x7ff8289c0000)
>>>> KERNELBASE.dll => /c/WINDOWS/System32/KERNELBASE.dll (0x7ff8280c0000)
>>>> cygplot-2.dll => /usr/bin/cygplot-2.dll (0x9f0000)
>>>> cygplot-2.dll => /usr/bin/cygplot-2.dll (0xbf0000)
>>>> cygplot-2.dll => /usr/bin/cygplot-2.dll (0xbf0000)
>>>> cygwin1.dll => /usr/bin/cygwin1.dll (0xd20000)
>>>> [...]
>>>> cygintl-8.dll => /usr/bin/cygintl-8.dll (0x1760000)
>>>> cygintl-8.dll => /usr/bin/cygintl-8.dll (0x1760000)
>>>> cygiconv-2.dll => /usr/bin/cygiconv-2.dll (0x1780000)
>>>> cygiconv-2.dll => /usr/bin/cygiconv-2.dll (0x1780000)

>>> can you check with
>>>    cat /proc/self/maps
>>> Usually I see

> Again nothing unusual.

>> There really does seem to be something peculiar about the graph program. I 
>> don't recall ever seeing this behavior with any other program.  Since it 
>> doesn't happen to everyone, it might depend on the Windows version.  Mine
>> is Windows 10 1909, Build 18363.959.
>>
>> I'm going to drop out of this discussion now.  I'm not a graph user, and I 
>> only jumped in to confirm that I could reproduce the problem, so that the
>> other posters wouldn't think the problem was just with their systems.

> My office machine (one of the problem machines) is running exactly the same
> version/build of Windows 10.

Office machines tend to run Enterprise builds which may be customized in many
ways unlike OEM/Retail/Home W10.

Check with `which -a graph` that some MS or other module is not being injected
into your search path.

-- 
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada

This email may be disturbing to some readers as it contains
too much technical detail. Reader discretion is advised.
[Data in IEC units and prefixes, physical quantities in SI.]