On 23.05.2020 10:48, Marco Atzeri wrote:
As I am blocked on the python update,
I would like to inform you about the issue
On i686 arch and only on that, in all versions
3.8.3 and 2.7.18 the sh.exe segfaults during configure
The logs on 3.8.3 and also rebuilding 3.8.2
---
checking for use_env... yes
configure: checking for device files
checking for /dev/ptmx... yes
make: *** No targets specified and no makefile found. Stop.
*** ERROR: make failed
---
on 2.7.18:
---
checking for /dev/ptmx... yes
checking for /dev/ptc... no
checking for %lld and %llu printf() format support... make: *** No
targets specified and no makefile found. Stop.
ESC[1;31m*** ERROR:ESC[0;0m make failed
15 [main] sh 6354 D:\cygwin32\bin\sh.exe: *** fatal error in
forked process - WFSO timed out after longjmp
80937 [main] sh 6354 cygwin_exception::open_stackdumpfile: Dumping
stack trace to sh.exe.stackdump
---
In one of the 3 cases no stack dump is produced, while for the other
two I have:
$ cat sh.exe.stackdump
Stack trace:
Frame Function Args
006DC078 61084F28 (, 77BE6DD2, 00777E68, 0077)
$ cat sh.exe.stackdump
Stack trace:
Frame Function Args
006DC438 61084F28 (0148, , , 0018)
End of stack trace
$ addr2line.exe -a 61084F28 -e /usr/bin/cygwin1.dll
0x61084f28
/usr/src/debug/cygwin-3.1.4-1/winsup/cygwin/fork.cc:650
As 3.8.2 was built before, it seems a recent change in Cygwin
is impacting the 32bit platform.
It seems I forgot the announce, but last build was 10th of April
https://cygwin.com/packages/x86/python38/python38-3.8.2-1
so I will try to bisect Cygwin to see when this problem arose.
Crossing finger
Marco
It seems I was lucky, current git does not have the problem
so I was able to build both 3.8.3 and 2.7.18
without any need for a bisect.
Give me some time for the packaging and I will update both
Regards
Marco
