Rtl function table may be partially corrupt

[jiahao]

julia> ; ls ~/nosuchfile
readline: Please submit a bug report with steps to reproduce this fault, and any error messages that follow (in their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x7ffba2a0bb96 -- RtlLeaveCriticalSection at ???: offset a2a0bb96
RtlLeaveCriticalSection at ???: offset a2a0bb96
RtlEnterCriticalSection at ???: offset a2a0b623
vfprintf at ???: offset a09e3161
vfprintf at ???: offset a09e30fe
jl_read_buffer at ???: offset 4104f6
jl_read_buffer at ???: offset 4120a9
anonymous at multi.jl:156
??? at ???: offset 23468f0
??? at ???: offset 42ea890
??? at ???: offset 3e61c40
ls: ~/nosuchfile: No such file or directory

Version 0.3.0-pre+705 (commit 1668d1b), built on MinGW-builds/MSYS (x86_64-w64-mingw32). This does not seem to be a problem in 0.2-release. Without ~, produces a process failed error without crashing REPL.

[ihnorton]

I don't think we support tilde expansion at all right now (see #1136 and #3919). Definitely shouldn't crash, regardless.

[vtjnash]

32-bit or 64-bit? curious mostly just because it generated a (mostly erroneous) backtrace.

also, git hashes are much more useful than version numbers

edit: also, this doesn't crash for me

[jiahao]

Updated with more details.

[jiahao]

It's interesting what works and what doesn't:

• ; ls * works
• ; ls .* works
• ; ls ~ works
• ; ls ~/* works
• ; ls ~/.* fails consistently
• ; ls /usr/* works
• ; ls /usr/.* fails consistently

so it looks like it's not limited to ~. (The tilde expansion is being handled by the shell, I think, not by Julia).

[vtjnash]

still doesn't crash for me.

unrelated observation: offset is getting truncated to an int before printing

[vtjnash]

@jiahao are you on IRC?

any idea who is printing readline:? does the various parts of the error message always get printed in the same order? I'm surprised ls: ~/nosuchfile: No such file or directory manages to get printed after it supposedly crashed.

[jiahao]

Ugh, apparently this manages to corrupt stdin. I tried to trap the bug in gdb but it manages to make gdb unresponsive.

[vtjnash]

This seems to actually be a dup of #5248.

Although I also fixed the issue reference by the replacement title of this issue in 3f423e3