Fatal Python error: Py_Initialize: can't initialize sys

[ljgabc]

Hi, I build Python-3.4.2 follow the build.sh, but run failed.

python python3.4.2
android 4.4.3

export PYTHONHOME=/data/Python-3.4.2

/data/Python-3.4.2/bin/python3

Fatal Python error: Py_Initialize: can't initialize sys
sys.excepthook is missing
object : ValueError('character U+7461642f is not in range [U+0000; U+10ffff]',)
type : ValueError
refcount: 1
address : 0x76cf7660
lost sys.stderr
[1] + Stopped (signal) python

[shizmob]

Hi!

It seems like it’s encountering some invalid bytestream very early on in the sys module initialisation, possibly related to an invalid filesystem path.
Judging from this code it could be a lot of things, but I suspect it’s one of the statements that start from here that causes the error.

Could I possibly get the output of running env in your Android shell, and would it be possible for you to hook up gdb to it and get a backtrace of the crash?
Alternatively, running arm-none-eabi-addr2line -e build/*/bin/python3 0x76cf7660 on the build computer could also help a lot to help me figure out where exactly the crash occurred.

Thanks!

[ljgabc]

Hi,
I recompiled the source and now it runs successfully on my board, so I cannot get the detail information again. Maybe it's my own mistake. Sorry for this, I'll close this issue.

[shizmob]

Glad it worked out, don't be afraid to re-open if the issue pops up again. :)

[michalliu]

1705     SET_SYS_FROM_STRING("executable",
1706                         PyUnicode_FromWideChar(
1707                                Py_GetProgramFullPath(), -1));
1708     SET_SYS_FROM_STRING("prefix",
1709                         PyUnicode_FromWideChar(Py_GetPrefix(), -1));
1710     SET_SYS_FROM_STRING("exec_prefix",
1711                         PyUnicode_FromWideChar(Py_GetExecPrefix(), -1));
1712     SET_SYS_FROM_STRING("base_prefix",
1713                         PyUnicode_FromWideChar(Py_GetPrefix(), -1));

It crashed at 1705. My error message is same as @ljgabc 's. It will happen once you set the PYTHONHOME environment variable.

[shizmob]

Does this also occur with @yan12125's fork? I haven't looked into this for a while, honestly. If it does, I'll be sure to check out the cause!