http://bugs.winehq.org/show_bug.cgi?id=34941
Bug #: 34941 Summary: Ruby interpreter hangs with OS-level thread concurrency error Product: Wine Version: 1.6.1 Platform: x86 OS/Version: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: ntdll AssignedTo: wine-bugs@winehq.org ReportedBy: jsnydr@gmail.com Classification: Unclassified
I just created a new 'version' entity for ruby in the appDB for ruby version 1.9.3 to report that it doesn't run. The recommended installer from http://rubyinstaller.org runs to completion and the installed 'ruby.exe' will report its version with `wine ruby -v`. But, if you try to get ruby to do anything (like install gems or run 'irb'), it becomes unresponsive such that it must be killed externally and begins generating the same error message every minute: 'err:ntdll:RtlpWaitForCriticalSection section 0x113848 "?" wait timed out in thread 0043, blocked by 0000, retrying (60 sec)'
http://bugs.winehq.org/show_bug.cgi?id=34941
jsnydr@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |jsnydr@gmail.com
http://bugs.winehq.org/show_bug.cgi?id=34941
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |download, source Status|UNCONFIRMED |NEW URL| |http://rubyinstaller.org Component|ntdll |-unknown Ever Confirmed|0 |1
--- Comment #1 from Austin English austinenglish@gmail.com 2013-11-18 20:00:52 CST --- Confirming.
austin@aw25 ~ $ sha1sum rubyinstaller-2.0.0-p247.exe 00632743bbec9872340bbe36884f1570ad3dc8ea rubyinstaller-2.0.0-p247.exe austin@aw25 ~ $ du -h rubyinstaller-2.0.0-p247.exe 17M rubyinstaller-2.0.0-p247.exe austin@aw25 ~ $ wine --version wine-1.7.6-112-geebc40f
http://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- URL|http://rubyinstaller.org |http://dl.bintray.com/onecl | |ick/rubyinstaller/rubyinsta | |ller-1.9.3-p448.exe CC| |focht@gmx.net Component|-unknown |msvcrt Summary|Ruby interpreter hangs with |Ruby 1.9.x interpreter |OS-level thread concurrency |hangs with OS-level thread |error |concurrency error (msvcrt | |internal semi-documented | |'ioinfo' structure layout | |mismatch)
--- Comment #2 from Anastasius Focht focht@gmx.net 2013-12-10 18:08:47 CST --- Hello folks,
confirming.
--- snip --- $ pwd /home/focht/.wine/drive_c/Ruby193/bin
$ WINEDEBUG=+tid,+seh,+relay wine ./ruby.exe >>log.txt 2>&1 ... 0024:Call msvcrt._setmode(00000003,00008000) ret=62d6642b 0024:Ret msvcrt._setmode() retval=00004000 ret=62d6642b 0024:Call msvcrt._get_osfhandle(00000003) ret=62e6f887 0024:Ret msvcrt._get_osfhandle() retval=0000005c ret=62e6f887 0024:Call msvcrt._get_osfhandle(00000003) ret=62e6f8ac 0024:Ret msvcrt._get_osfhandle() retval=0000005c ret=62e6f8ac 0024:Call KERNEL32.PeekConsoleInputA(0000005c,0061f088,00000001,0061f09c) ret=62e6f8d1 0024:Ret KERNEL32.PeekConsoleInputA() retval=00000000 ret=62e6f8d1 0024:Call msvcrt.malloc(00002000) ret=62d55130 0024:Call ntdll.RtlAllocateHeap(00110000,00000000,00002000) ret=7ec45a47 0024:Ret ntdll.RtlAllocateHeap() retval=001fa2d0 ret=7ec45a47 0024:Ret msvcrt.malloc() retval=001fa2d0 ret=62d55130 0024:trace:seh:__regs_MSVCRT__setjmp buf=0x133918 ebx=00133850 esi=001338bc edi=00000000 ebp=0061f018 esp=0061effc eip=62e5f8da frame=0061fed8 0024:Call KERNEL32.ReleaseMutex(00000050) ret=62e631d3 0024:Ret KERNEL32.ReleaseMutex() retval=00000001 ret=62e631d3 0024:Call msvcrt._get_osfhandle(00000003) ret=62e76c91 0024:Ret msvcrt._get_osfhandle() retval=0000005c ret=62e76c91 0024:Call msvcrt._get_osfhandle(00000003) ret=62e76cb6 0024:Ret msvcrt._get_osfhandle() retval=0000005c ret=62e76cb6 0025:Ret KERNEL32.WaitForSingleObject() retval=00000102 ret=62e5d039 0025:Call KERNEL32.WaitForSingleObject(00000054,00000064) ret=62e5d039 0025:Ret KERNEL32.WaitForSingleObject() retval=00000102 ret=62e5d039 ... 0025:Call KERNEL32.WaitForSingleObject(00000054,00000064) ret=62e5d039 0024:err:ntdll:RtlpWaitForCriticalSection section 0x122108 "?" wait timed out in thread 0024, blocked by 0000, retrying (60 sec) 0025:Ret KERNEL32.WaitForSingleObject() retval=00000102 ret=62e5d039 --- snip ---
Tid 0025 seems harmless. Looks like a fast running ruby 'timer' thread which peeks every 100ms for some work and executes 'tasklets' if needed.
Tid 0024 is the culprit here. Using the version info along with some debugging/disassembly I found the code which most likely resembles the problem/trace log -> rb_w32_read():
Source: http://svn.ruby-lang.org/cgi-bin/viewvc.cgi/branches/ruby_1_9_3/win32/win32....
(binary 'msvcrt-ruby191.dll')
--- snip --- 5183 #undef read 5184 ssize_t 5185 rb_w32_read(int fd, void *buf, size_t size) 5186 { 5187 SOCKET sock = TO_SOCKET(fd); 5188 DWORD read; 5189 DWORD wait; 5190 DWORD err; 5191 size_t len; 5192 size_t ret; 5193 OVERLAPPED ol, *pol = NULL; 5194 BOOL isconsole; 5195 BOOL islineinput = FALSE; 5196 int start = 0; 5197 5198 if (is_socket(sock)) 5199 return rb_w32_recv(fd, buf, size, 0); 5200 5201 // validate fd by using _get_osfhandle() because we cannot access _nhandle 5202 if (_get_osfhandle(fd) == -1) { 5203 return -1; 5204 } 5205 5206 if (_osfile(fd) & FTEXT) { 5207 return _read(fd, buf, size); 5208 } 5209 5210 MTHREAD_ONLY(EnterCriticalSection(&(_pioinfo(fd)->lock))); 5211 ... --- snip ---
Line 5210 seems to be the matching source line to EnterCriticalSection -> RtlpWaitForCriticalSection (see timeout message with the invalid blocking tid)
The problem seems to be that Ruby's win32 wrapper makes direct use of msvcrt internal data structures, making assumptions about the layout.
--- snip --- ... 2021 typedef struct { 2022 intptr_t osfhnd; /* underlying OS file HANDLE */ 2023 char osfile; /* attributes of file (e.g., open in text mode?) */ 2024 char pipech; /* one char buffer for handles opened on pipes */ 2025 #ifdef MSVCRT_THREADS 2026 int lockinitflag; 2027 CRITICAL_SECTION lock; 2028 #endif 2029 #if RT_VER >= 80 2030 char textmode; 2031 char pipech2[2]; 2032 #endif 2033 } ioinfo; 2034 2035 #if !defined _CRTIMP || defined __MINGW32__ 2036 #undef _CRTIMP 2037 #define _CRTIMP __declspec(dllimport) 2038 #endif 2039 2040 #if !defined(__BORLANDC__) 2041 EXTERN_C _CRTIMP ioinfo * __pioinfo[]; 2042 2043 #define IOINFO_L2E 5 2044 #define IOINFO_ARRAY_ELTS (1 << IOINFO_L2E) 2045 #define _pioinfo(i) ((ioinfo*)((char*)(__pioinfo[i >> IOINFO_L2E]) + (i & (IOINFO_ARRAY_ELTS - 1)) * (sizeof(ioinfo) + pioinfo_extra))) 2046 #define _osfhnd(i) (_pioinfo(i)->osfhnd) 2047 #define _osfile(i) (_pioinfo(i)->osfile) 2048 #define _pipech(i) (_pioinfo(i)->pipech) 2049 2050 #if RT_VER >= 80 2051 static size_t pioinfo_extra = 0; /* workaround for VC++8 SP1 */ ... --- snip ---
If we compare it to Wine's ioinfo, which is exported through __pioinfo (internally MSVCRT___pioinfo) array:
Source: http://source.winehq.org/git/wine.git/blob/3bb2c11757156223be0f8a22f3b377b06...
--- snip --- 84 /* ioinfo structure size is different in msvcrXX.dll's */ 85 typedef struct { 86 HANDLE handle; 87 unsigned char wxflag; 88 char lookahead[3]; 89 int exflag; 90 CRITICAL_SECTION crit; 91 } ioinfo; --- snip ---
This obviously can't work.
Maybe Wine's ioinfo structure layout could be made more compatible with Ruby's ioinfo definition, matching MS CRT.
There are various other open source projects which directly access this 'semi-documented' internal structure (python, stlport, cygwin, msys ...). Making Wine more compatible would help other python, cygwin, msys related bugs.
$ sha1sum rubyinstaller-1.9.3-p448.exe 728dae18b46611a28b2ae542e68eb0d22c66a087 rubyinstaller-1.9.3-p448.exe
$ du -sh rubyinstaller-1.9.3-p448.exe 15M rubyinstaller-1.9.3-p448.exe
$ wine --version wine-1.7.8-135-gfaa355b
Regards
http://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Blocks| |13953
http://bugs.winehq.org/show_bug.cgi?id=34941
Piotr Caban piotr.caban@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |piotr.caban@gmail.com
--- Comment #3 from Piotr Caban piotr.caban@gmail.com --- In this case ioinfo structure layout is correct (ruby is using msvcrt.dll, _RT_VER is not defined during compilation). If it uses msvcr80.dll it will not work because of the size mismatch.
The problem is that we currently don't use this critical sections and don't initialize them.
http://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |maiktapwagner@aol.com
--- Comment #4 from Anastasius Focht focht@gmx.net --- *** Bug 35832 has been marked as a duplicate of this bug. ***
http://bugs.winehq.org/show_bug.cgi?id=34941
Patrick Hanevold patrick.hanevold@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |patrick.hanevold@gmail.com
http://bugs.winehq.org/show_bug.cgi?id=34941
Vijay Kamuju infyquest@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |infyquest@gmail.com
--- Comment #5 from Vijay Kamuju infyquest@gmail.com --- I think a patch for fixing the ioinfo has been committed, please test it with latest wine git
The commit is http://source.winehq.org/git/wine.git/commitdiff/42e800fe796665d914f8ab7f934...
http://bugs.winehq.org/show_bug.cgi?id=34941
--- Comment #6 from Anastasius Focht focht@gmx.net --- Hello Vijay,
--- quote --- I think a patch for fixing the ioinfo has been committed, please test it with latest wine git --- quote ---
the version-independent part of the 'ioinfo' structure is what matters here - and that is still unchanged (that is before '#if _MSVCR_VER') - so no.
http://source.winehq.org/git/wine.git/blob/42e800fe796665d914f8ab7f93495e0c1...
Regards
http://bugs.winehq.org/show_bug.cgi?id=34941
Richard Neish richardn@richardneish.org changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |richardn@richardneish.org
https://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |u.schutz@bluewin.ch
--- Comment #7 from Anastasius Focht focht@gmx.net --- *** Bug 35870 has been marked as a duplicate of this bug. ***
https://bugs.winehq.org/show_bug.cgi?id=34941
Ryan Tandy ryan@nardis.ca changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |ryan@nardis.ca
https://bugs.winehq.org/show_bug.cgi?id=34941
s.wanabe@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |s.wanabe@gmail.com
--- Comment #8 from s.wanabe@gmail.com --- I have tested ruby-2.0.0-p576-i386-mingw32 on latest wine 1.7.28. It has not worked yet.
I guess msvcrt_set_fd() needs to initialize fdinfo->crit because win32.c (in ruby) includes "EnterCriticalSection(&(_pioinfo(fd)->lock)".
I confirmed that "wine ruby.exe -e 'p 1'" worked fine with below patch.
diff --git a/dlls/msvcrt/file.c b/dlls/msvcrt/file.c index 7ce8b59..a7f3a5c 100644 --- a/dlls/msvcrt/file.c +++ b/dlls/msvcrt/file.c @@ -380,7 +380,8 @@ static int msvcrt_set_fd(HANDLE hand, int flag, int fd) fdinfo->lookahead[0] = '\n'; fdinfo->lookahead[1] = '\n'; fdinfo->lookahead[2] = '\n'; - fdinfo->exflag = 0; + InitializeCriticalSection(&fdinfo->crit); + fdinfo->exflag = EF_CRIT_INIT;
/* locate next free slot */ if (fd == MSVCRT_fdstart && fd == MSVCRT_fdend)
https://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |patch Summary|Ruby 1.9.x interpreter |Ruby 1.9.x interpreter |hangs with OS-level thread |hangs with OS-level thread |concurrency error (msvcrt |concurrency error (critical |internal semi-documented |section not initialized in |'ioinfo' structure layout |msvcrt internal |mismatch) |semi-documented 'ioinfo' | |struct)
--- Comment #9 from Anastasius Focht focht@gmx.net --- Hello folks,
refining summary and adding 'patch' keyword.
Regards
https://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Fixed by SHA1| |0c4cb6ce5d411a60ec11e52d648 | |1c4148344d067 Status|NEW |RESOLVED Resolution|--- |FIXED
--- Comment #10 from Anastasius Focht focht@gmx.net --- Hello folks,
this is fixed by commit http://source.winehq.org/git/wine.git/commitdiff/0c4cb6ce5d411a60ec11e52d648...
Thanks Piotr
--- snip --- $ wine ./ruby gem list
*** LOCAL GEMS ***
bigdecimal (1.1.0) io-console (0.3) json (1.5.5) minitest (2.5.1) rake (0.9.2.2) rdoc (3.9.5) --- snip ---
Fetching and unpacking Gems test:
--- snip --- $ pwd /home/focht/.wine/drive_c/Ruby193/bin
$ wine ./ruby gem fetch malice Fetching: malice-13.gem (100%) Downloaded malice-13
$ wine ./ruby gem unpack malice Fetching: malice-13.gem (100%) Unpacked gem: 'C:/Ruby193/bin/malice-13'
$ more malice-13/README
Malice v. 13
DESCRIPTION
A small, malicious library.
Use at your own risk. No warranty of fitness! Copyright 2007 Cloudburst, LLC. Licensed under the AFL 3.0; see the included LICENSE file. --- snip ---
For interactive Ruby shell (irb) to be useful you need to start with 'wineconsole'.
Regards
https://bugs.winehq.org/show_bug.cgi?id=34941
--- Comment #11 from Piotr Caban piotr.caban@gmail.com --- The application may use the critical section now but it's not locked when it should be. There's still quite a lot to be done before it's fully implemented. Anyway I'm OK with both marking the bug as fixed or leaving it as not yet fixed.
https://bugs.winehq.org/show_bug.cgi?id=34941
Alexandre Julliard julliard@winehq.org changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #12 from Alexandre Julliard julliard@winehq.org --- Closing bugs fixed in 1.7.31.
https://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |aerilius@googlemail.com
--- Comment #13 from Anastasius Focht focht@gmx.net --- *** Bug 38023 has been marked as a duplicate of this bug. ***
https://bugs.winehq.org/show_bug.cgi?id=34941
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- URL|http://dl.bintray.com/onecl |https://web.archive.org/web |ick/rubyinstaller/rubyinsta |/20201205015427/http://dl.b |ller-1.9.3-p448.exe |intray.com/oneclick/rubyins | |taller/rubyinstaller-1.9.3- | |p448.exe?direct
-
wine-bugs＠winehq.org -
WineHQ Bugzilla