[Bug 16948] New: tmpfile() write, seek(0) and read does not return same data

Summary: tmpfile() write, seek(0) and read does not return same data Product: Wine Version: unspecified Platform: All OS/Version: Linux Status: UNCONFIRMED Severity: enhancement Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: lkcl@lkcl.net

ok this is tested from python (import os) so it _could_ be done a little simpler - like... using c. didn't want to forget about it so am raising it here.

this is the mode opening a tmpfile:

fixme:msvcrt:MSVCRT__sopen : pmode 0x1464e0 ignored

http://bugs.python.org/issue4956 $ ./python.exe -i Python 2.5.2 (r252:60911, Jan 15 2009, 23:28:55) [gcc] on win32 Type "help", "copyright", "credits" or "license" for more information.

...
...
import os f = os.tmpfile() f.name

'<tmpfile>'

...
...
f.write("hello\n") f.flush() f

...
...
f.tell()

...
...
f.seek(0) f.tell()

...
...
f.read()

it's as if it's impossible to read from the tmpfile. which kinda defeats the object of having one :)

c:\windows\temp exists but c:\windows\tmp does not.

-- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email Do not reply to this email, post in Bugzilla using the above URL to reply. ------- You are receiving this mail because: ------- You are watching all bug changes.

Show replies by date

wine-bugs＠winehq.org

16 Jan 16 Jan

3:22 a.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

Dan Kegel dank@kegel.com changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |dank@kegel.com

--- Comment #1 from Dan Kegel dank@kegel.com 2009-01-15 21:22:38 --- You forgot to mention that, according to that bug report, you need msys to trigger the bug, too.

Can you provide a linux shell script to reproduce the problem from start to finish?

wine-bugs＠winehq.org

4:10 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

--- Comment #2 from Luke Kenneth Casson Leighton lkcl@lkcl.net 2009-01-16 10:10:13 --- hiya dan, actually it's much simpler - and much worse! :) when running wine z:/src/winebug/a.exe i get the attached segfault. when running the same program under msys, no segfault occurs, but the reported bug (seek works but read fails) can be duplicated. gcc -g -g test.c a.exe also attached (otherwise you have to install msys and mingw under wine).

wine-bugs＠winehq.org

4:11 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

--- Comment #3 from Luke Kenneth Casson Leighton lkcl@lkcl.net 2009-01-16 10:11:03 --- Created an attachment (id=18730) --> (http://bugs.winehq.org/attachment.cgi?id=18730) source code test case

wine-bugs＠winehq.org

4:11 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

--- Comment #4 from Luke Kenneth Casson Leighton lkcl@lkcl.net 2009-01-16 10:11:26 --- Created an attachment (id=18731) --> (http://bugs.winehq.org/attachment.cgi?id=18731) precompiled test case (gcc -g -g test.c)

wine-bugs＠winehq.org

4:11 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

--- Comment #5 from Luke Kenneth Casson Leighton lkcl@lkcl.net 2009-01-16 10:11:45 --- Created an attachment (id=18732) --> (http://bugs.winehq.org/attachment.cgi?id=18732) segfault of wine ./a.exe

wine-bugs＠winehq.org

5:19 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

Austin English austinenglish@gmail.com changed:

What |Removed |Added ---------------------------------------------------------------------------- Keywords| |download, testcase

wine-bugs＠winehq.org

5:29 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

Vincent Povirk madewokherd@gmail.com changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |madewokherd@gmail.com

wine-bugs＠winehq.org

5:45 p.m.

New subject: [Bug 16948] tmpfile() write, seek(0) and read does not return same data

Juan Lang juan_lang@yahoo.com changed:

wine-bugs＠winehq.org

5:56 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Juan Lang juan_lang@yahoo.com changed:

--- Comment #6 from Juan Lang juan_lang@yahoo.com 2009-01-16 11:56:39 --- Confirming, though I compiled like so: winegcc -g test.c -lmsvcrt

Omitting -lmsvcrt didn't produce the crash, as it used gcc's libc instead.

Here's a snip from my backtrace (which has a little more info): =>0 0x601cf01e MSVCRT_fwrite+0x2e(ptr=0x6019f7d6, size=5, nmemb=1, file=(nil)) [/home/juan/src/wine/dlls/msvcrt/file.c:2403] in msvcrt (0x0032fe88) 1 0x6019f583 main+0x5f() [/home/juan/foo/tmp/test.c:9] in a.out (0x0032fed8) 2 0x6019f6d8 __wine_spec_exe_entry+0x88(peb=0x7ffdf000) [/home/juan/src/wine/dlls/winecrt0/exe_entry.c:36] in a.out (0x0032ff08) 3 0x7b8762a8 start_process+0x98(arg=(nil)) [/home/juan/src/wine/dlls/kernel32/process.c:904] in kernel32 (0x0032ffe8) 4 0x6413a697 wine_switch_to_stack+0x17() in libwine.so.1 (0x00000000) 0x601cf01e MSVCRT_fwrite+0x2e [/home/juan/src/wine/dlls/msvcrt/file.c:2403] in msvcrt: movl 0x4(%edi),%eax 2403 if(file->_cnt) {

The problem is tmpfile() is returning NULL. Here's a snip from a +relay log: 0009:Call msvcrt.tmpfile() ret=60321546 0009:Call KERNEL32.GetFileAttributesA(6038cc60 "\s8.") ret=6034bf27 0009:Ret KERNEL32.GetFileAttributesA() retval=ffffffff ret=6034bf27 fixme:msvcrt:MSVCRT__sopen : pmode 0x7bca91a1 ignored 0009:Call KERNEL32.CreateFileA(6038cc60 "\s8.",c0010000,00000007,0032fdb0,00000004,04000080,00000000) ret=6034e85a 0009:Ret KERNEL32.CreateFileA() retval=ffffffff ret=6034e85a

Note the path: "\s8.". tmpnam() returns a filename in the root directory of whatever the current drive is: p = s + sprintf(s, "\s%s.", tmpstr);

The trouble comes about when running the test program from a Unix path. Copying it to somewhere in the c: drive causes it to succeed.

I'm guessing tmpnam() needs to be adjusted to return a file in %TMP%, based on its description on MSDN, but without test cases to confirm it I'm not sure.

wine-bugs＠winehq.org

5:57 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Juan Lang juan_lang@yahoo.com changed:

What |Removed |Added ---------------------------------------------------------------------------- Version|unspecified |1.1.13

--- Comment #7 from Juan Lang juan_lang@yahoo.com 2009-01-16 11:57:47 --- Forgot to specify the Wine version.

wine-bugs＠winehq.org

6:03 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

--- Comment #8 from Juan Lang juan_lang@yahoo.com 2009-01-16 12:03:57 --- Created an attachment (id=18733) --> (http://bugs.winehq.org/attachment.cgi?id=18733) Patch

This patch fixes it, though a little testing wouldn't hurt.

wine-bugs＠winehq.org

6:04 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Juan Lang juan_lang@yahoo.com changed:

What |Removed |Added ---------------------------------------------------------------------------- Keywords| |patch

wine-bugs＠winehq.org

6:18 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Juan Lang juan_lang@yahoo.com changed:

--- Comment #9 from Juan Lang juan_lang@yahoo.com 2009-01-16 12:18:33 --- Hmm, according to tmpfile()'s description on MSDN, "The temporary file is created in the root directory." Thus the patch is incorrect, and the problem is that z:\ is not writable by you.

Workaround: run from a drive whose root is writable, e.g. c:

wine-bugs＠winehq.org

6:19 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Juan Lang juan_lang@yahoo.com changed:

wine-bugs＠winehq.org

6:25 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

--- Comment #10 from Juan Lang juan_lang@yahoo.com 2009-01-16 12:25:57 --- For further reference, glibc works around this "feature" of msvcrt. See the following discussion of it, and their workaround: http://www.mail-archive.com/bug-gnulib@gnu.org/msg05721.html

wine-bugs＠winehq.org

6:53 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Dan Kegel dank@kegel.com changed:

--- Comment #11 from Dan Kegel dank@kegel.com 2009-01-16 12:53:49 --- gnulib != glibc, but the link's interesting anyway.

I think we should make this work for the common case where z:/ is not writable somehow, so reopening.

wine-bugs＠winehq.org

20 Jan 20 Jan

5:08 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Dmitry Timoshkov dmitry@codeweavers.com changed:

What |Removed |Added ---------------------------------------------------------------------------- Platform|All |Other

wine-bugs＠winehq.org

5 Aug 5 Aug

1:38 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

--- Comment #12 from Austin English austinenglish@gmail.com 2010-08-05 08:38:56 --- With the precompiled exe, in wine-1.3.0-167-g65d1c99

putting it in ~/.wine/drive_c gives

austin@laptop:~/.wine/drive_c$ wine a.exe fixme:msvcrt:MSVCRT__sopen : pmode 0x60fdc8 ignored filepos: 5 filepos: 0 buffer: �/h�@filepos: 0

and in $HOME, I get a crash:

Backtrace: =>0 0x6858b4a7 MSVCRT_fwrite+0x27(ptr=0x403000, size=0x0005, nmemb=0x0001, file=(nil)) [/home/austin/wine-git/dlls/msvcrt/file.c:2564] in msvcrt (0x0060fe48) 1 0x00401345 main+0x54(argc=0x0001, argv=0x1103a0) [C:/src/winebug-16948/test.c:7] in a (0x0060fe80) 2 0x00401298 in a (+0x1297) (0x0060fe90)

native msvcrt crashes from $HOME as well, though...however, the drive_c results are different:

austin@laptop:~/.wine/drive_c$ wine a.exe filepos: 5 filepos: 0 buffer: hello(filepos: 5

(notice the unreadable characters instead of 'hello').

wine-bugs＠winehq.org

21 Sep 21 Sep

2:54 a.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Bruno Jesus 00cpxxx@gmail.com changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |00cpxxx@gmail.com

--- Comment #13 from Bruno Jesus 00cpxxx@gmail.com 2011-09-20 21:54:17 CDT --- Still present in 1.3.28.

wine-bugs＠winehq.org

16 Oct 16 Oct

3:39 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Piotr Caban piotr.caban@gmail.com changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |piotr.caban@gmail.com

--- Comment #14 from Piotr Caban piotr.caban@gmail.com 2012-10-16 10:39:48 CDT --- tmpfile fails when root directory is read-only on windows unless VirtualStore is enabled.

What was tested: - tmpfile on read-only file system - tmpfile fails - tmpfile on c: drive (user without privileges to write directly to c:\ directory), the file was created in VirtualStore, function has succeeded

I've sent a patch that fixes attached test when tmpfile succeeds (http://source.winehq.org/patches/data/91032).

wine-bugs＠winehq.org

15 Mar 15 Mar

10:26 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Sebastian Lackner sebastian@fds-team.de changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |sebastian@fds-team.de

-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

wine-bugs＠winehq.org

23 Jul 23 Jul

12:13 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Ken Sharp imwellcushtymelike@gmail.com changed:

--- Comment #15 from Ken Sharp imwellcushtymelike@gmail.com --- Was the patch ever accepted?

-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

wine-bugs＠winehq.org

1:21 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

--- Comment #16 from Piotr Caban piotr.caban@gmail.com --- (In reply to Ken Sharp from comment #15)

Was the patch ever accepted?

Yes, the patch I've mentioned was committed.

It's not clear for me if there's a bug. Currently wine behaves the same as Windows XP. On the other hand it would be nice to have support for VirtualStore so the function may work if root path is read-only.

-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

wine-bugs＠winehq.org

6 Jun 6 Jun

8:37 a.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Piotr Caban piotr.caban@gmail.com changed:

--- Comment #17 from Piotr Caban piotr.caban@gmail.com --- Starting from msvcr110 native creates the file in temp directory. I have changed wine's implementation to do the same for all msvcr* versions. Marking the bug as fixed.

-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

wine-bugs＠winehq.org

9 Jun 9 Jun

5:45 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Alexandre Julliard julliard@winehq.org changed:

What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED

--- Comment #18 from Alexandre Julliard julliard@winehq.org --- Closing bugs fixed in 2.10.

-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

wine-bugs＠winehq.org

9 Aug 9 Aug

11:24 a.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path

Michael Stefaniuc mstefani@winehq.org changed:

What |Removed |Added ---------------------------------------------------------------------------- Target Milestone|--- |2.0.x

-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

wine-bugs＠winehq.org

20 Oct 20 Oct

12:41 p.m.

New subject: [Bug 16948] tmpfile() fails when run from Unix path