October 2016 - Wine-Bugs - list.winehq.org

[Bug 41437] New: Implement a 64-bit version of __std_type_info_hash
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=41437 Bug ID: 41437 Summary: Implement a 64-bit version of __std_type_info_hash Product: Wine Version: unspecified Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: msvcrt Assignee: wine-bugs(a)winehq.org Reporter: cerebro.alexiel(a)gmail.com Distribution: --- Two weeks ago, I saw commit 4931e6f92bc7e0c229a057ebf2e000f8f5aa1edd that (partially) implements __std_type_info_hash function. Code was simple but the 64-bit version was not implemented as it was using "different constants". I thought: "Findind a coefficient only requires time and patience, let's do it!". I had to find a value (noted as <?> here) that is such that (0xcbf29ce44fd0bfc1 ^ 0x1) * <?> = 0xaf63bc4c29620a60 or 0xcbf29ce44fd0bfc0 * <?> = 0xaf63bc4c29620a60. I started with a naive implementation computing all <?> values from 0 to UINT64_MAX. Was very long and very slow. I optimized it using OpenMP (#pragma omp parallel) causing all my CPUs to work at 100%. Still very slow and the topmost bit would be tested in ages. I then had the idea to use a (pseudo-)random non-repeating number generator. This way, all numbers would be tested in (pseudo-)random order and equally distributed on the 64-bit range. Plus, with the known seed and index, it would be posible to stop and continue the computation at the point it stopped. I chose Blowfish algorithm and let it work for about a week and then realized I may need a better solution. Finding the 32-bit <?> value was very quick (less than a minute) but a 64-bit value is way larger and testing every bit takes a really long time. Normally, if I have something like 5 * <?> = 15, I can do 15/5 = 3 = <?>. Here, we abuse overflow so, mathematically it's more (0xcbf29ce44fd0bfc0 * <?>) mod 2^64 = 0xaf63bc4c29620a60. I discovered we can actually do "Modular multiplicative inverse" to find an "inverse" of a number. <¿> such as (<¿> * a) mod x = 1 (a * <?>) mod x = b --> <?> mod x = b * <¿> I found a C++ implementation that finds the inverse of a number modulo another. I started with the 32-bit version to test it (no mod 2^32 to make it concise): (0x811c9dc5 ^ 0x1) * <?> = 0x040c5b8c 0x811c9dc4 * <?> = 0x040c5b8c 0x100000000=2^32 modInverse(0x811c9dc4, 0x100000000) -> no inverse because 0x811c9dc4 is even Anyway, let's do it with 0x2: (0x811c9dc5 ^ 0x2) * <?> = 0x070c6045 0x811c9dc7 * <?> = 0x070c6045 0x811c9dc7 inverse is 0xdce213f7 in other words, (0x811c9dc7 * 0xdce213f7) mod 0x100000000 = 1 so 0x070c6045 * 0xdce213f7 = <?> = 0x1000193 we can verify that 0x811c9dc7 * 0x1000193 = 0x70c6045 and that 0x811c9dc4 * 0x1000193 = 0x40c5b8c -> we have our 32-bit coefficient. Note that 0x100000000 requires 33 bits so it's ok on my 64-bit computer but as you may have guessed, 0x10000000000000000 (2^64) is 63 bits. I tried using uint128_t but didn't work. I then used GMP library (The GNU Multiple Precision Arithmetic Library) which lets me work on numbers as big as I have memory. Let's do it again for the 64-bit version (mox 2^64 omitted): (0xcbf29ce44fd0bfc1 ^ 0x1) * <?> = 0xaf63bc4c29620a60 0xcbf29ce44fd0bfc0 * <?> = 0xaf63bc4c29620a60 0xcbf29ce44fd0bfc0 has no inverse because 0xcbf29ce44fd0bfc0 is even (0xcbf29ce44fd0bfc1 ^ 0x2) * <?> = 0xaf63bf4c29620409 0xcbf29ce44fd0bfc3 * <?> = 0xaf63bf4c29620409 0xcbf29ce44fd0bfc3 inverse is 0xdbab92123bd8a8eb in other words, (0xcbf29ce44fd0bfc3 * 0xdbab92123bd8a8eb) mod 0x10000000000000000 = 1 so 0xaf63bf4c29620409 * 0xdbab92123bd8a8eb = <?> = 0xa01b8255ca379c43 we can verify that 0xcbf29ce44fd0bfc3 * 0xa01b8255ca379c43 = 0xaf63bf4c29620409 BUT 0xcbf29ce44fd0bfc0 * 0xa01b8255ca379c43 != 0xaf63bc4c29620a60 I tested with 0x2, 0x4, 0x42, inverse is ok but <?> was always different and wasn't working: I was missing something. I googled 0x1000193 and found out that it's a well-known number used in FNV-1 and FNV-1a hash algorithms (https://en.wikipedia.org/wiki/Fowler%E2%80%93Noll%E2%80%93Vo_hash_function) Ideal 64-bit <?> (called FNV_prime) is 0xcbf29ce484222325 which slightly differs from our 0xcbf29ce44fd0bfc1. Since only the low word is changed, I guessed it was a XOR operation on the low part. I tested 0xcbf29ce484222325 ^ i == 0xcbf29ce44fd0bfc1 in a loop and found i was 0xcbf29ce4. I retested with other values et voilà, I found how it worked! I just submitted a patch to fix this bug. I posted a detailed explanation here to prove I didn't disassembled the native DLL and maybe to tell people it's not that hard to do useful things in wine. By the way, it's a fun way to learn things! -- 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.

1 3

[Bug 41432] New: Keyboard input code display program ( Japanese VB6 application) doesn't run.
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=41432 Bug ID: 41432 Summary: Keyboard input code display program (Japanese VB6 application) doesn't run. Product: Wine Version: 1.9.18 Hardware: x86-64 URL: http://www.technoveins.co.jp/utility/keyindisp/index.h tm OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: kernel32 Assignee: wine-bugs(a)winehq.org Reporter: sagawa.aki+winebugs(a)gmail.com Distribution: --- Keyboard input code display program is a Japanese VB6 application. It shows keyboard scan code information etc. on native (Windows). However, on wine environment, it shows insufficient memory error (メモリが不足しています。終了します。) and doesn't run. -- 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.

1 5

[Bug 41455] New: Failed assertion in ME_SplitByBacktracking causes crash at end of Jupiter 2011 installer
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=41455 Bug ID: 41455 Summary: Failed assertion in ME_SplitByBacktracking causes crash at end of Jupiter 2011 installer Product: Wine Version: 1.9.20 Hardware: x86-64 URL: http://pity.elfin.pl/pliki/Jupiter/Jupiter2011/std/1.0 .5.6/Jupiter2011_std_1.0.5.exe OS: Linux Status: NEW Keywords: download Severity: normal Priority: P2 Component: richedit Assignee: wine-bugs(a)winehq.org Reporter: frederic.delanoy(a)gmail.com Blocks: 30522 Distribution: Ubuntu Created attachment 55823 --> https://bugs.winehq.org/attachment.cgi?id=55823 Crash backtrace At the end of the installer, one gets a crash from failed assertion Tested with wine-1.9.20-59-g7756d6b ; Ubuntu 14.04.5 SHA1 3f037f27bbb9654307723f0ac89b2aa39ba21a4c Jupiter2011_std_1.0.5.exe Log only shows: fixme:shell:SHAutoComplete stub wrap.c:500: ME_SplitByBacktracking: Assertion `i<len' failed. wine: Assertion failed at address 0xf777b440 (thread 0009), starting debugger... -- 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.

1 7

[Bug 39471] New: Elminage Gothic freezes after clicking on "New Game"
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=39471 Bug ID: 39471 Summary: Elminage Gothic freezes after clicking on "New Game" Product: Wine Version: 1.7.53 Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: blocker Priority: P2 Component: -unknown Assignee: wine-bugs(a)winehq.org Reporter: fjfrackiewicz(a)gmail.com Distribution: --- Created attachment 52601 --> https://bugs.winehq.org/attachment.cgi?id=52601 Terminal output for "wine Elminage.exe" I am using the GOG version of Elminage Gothic on 32-bit Wine 1.7.53 compiled from source. This issue is reproducible 100% of the time. Elminage Gothic has two movies it plays: One movie is the Ghostlight logo and then it plays an intro to the game once you click on "New Game". The game freezes when attempting to play the second movie. You can hear sound play in the background and all but the game is locked up completely. I've attached the terminal output from when you run the game via the "wine Elminage.exe" command in terminal. I have found a temporary workaround until this issue is somehow fixed: Rename the two movie files so that the intro shows up in the logo's place. This involves renaming "logo.wmv" to "intro.wmv" and "intro.wmv" to "logo.wmv". After that, you need to start pressing the Escape button just as you click on "New Game" to skip the second movie and get into the game proper. -- 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.

1 42

[Bug 39263] New: RichEdit should hide selection when it is unfocused
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=39263 Bug ID: 39263 Summary: RichEdit should hide selection when it is unfocused Product: Wine Version: 1.7.44 Hardware: x86 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: richedit Assignee: wine-bugs(a)winehq.org Reporter: vvort(a)yandex.ru Distribution: --- Created attachment 52335 --> https://bugs.winehq.org/attachment.cgi?id=52335 Test exe When riched20 control do not have input focus, it's text selection must be hidden. (Unless ES_NOHIDESEL is in effect) Test files (exe and sources) and screenshots are attached. This bug produces graphical glitch in ReactOS Application Manager. -- 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.

1 9

[Bug 40422] New: Illustrator CS6: hangs when on ESC key press while Text tool is being used
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=40422 Bug ID: 40422 Summary: Illustrator CS6: hangs when on ESC key press while Text tool is being used Product: Wine Version: 1.9.5 Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown Assignee: wine-bugs(a)winehq.org Reporter: vlad-mbx(a)ya.ru Distribution: --- Created attachment 54161 --> https://bugs.winehq.org/attachment.cgi?id=54161 wine debug output Illustrator CS6 hangs if I press Esc key while using Text tool. I can, however, switch to any other tool without problems using toolbar buttons. I have attached an output of "WINEDEBUG=+relay wine Illustrator.exe". The recording was started about 5 seconds before the hang was reproduced; I can provide a larger log file. The last API call GetAsyncKeyState() is repeated until Illustrator is forcibly closed. -- 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.

1 10

[Bug 41420] New: Regression bug: bricscad crashes when copy or cut objects to the clipboard
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=41420 Bug ID: 41420 Summary: Regression bug: bricscad crashes when copy or cut objects to the clipboard Product: Wine Version: 1.9.19 Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: major Priority: P2 Component: -unknown Assignee: wine-bugs(a)winehq.org Reporter: tolikk91(a)yandex.ru Distribution: --- Created attachment 55791 --> https://bugs.winehq.org/attachment.cgi?id=55791 wine's backtrace bricscad crashes when copy or cut objects to the clipboard. In wine version 1.9.18 of this error has not been. -- 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.

1 5

[Bug 22317] New: Motocross madness demo has wrongly projected smoke
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

http://bugs.winehq.org/show_bug.cgi?id=22317 Summary: Motocross madness demo has wrongly projected smoke Product: Wine Version: 1.1.42 Platform: x86 URL: http://www.microsoft.com/games/Motocross/downloads.htm OS/Version: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown AssignedTo: wine-bugs(a)winehq.org ReportedBy: wylda(a)volny.cz Created an attachment (id=27294) --> (http://bugs.winehq.org/attachment.cgi?id=27294) Smoke in Wine vs. WinXP Motocross madness demo has wrongly projected smoke. Normally it should come from behind the bike, but instead it comes from center of the screen, thus it becomes unplayable (key F9 turns this graphical effect off). Tested since 1.0-rc4 till now, but it isn't a regression. (Versions 1.1.18 and lower needs to apply commit fc87182d191a2b3daabf260d1ad4b12b157e3cd1 to be able to run mcm.exe) Demo is quite small: 19MB -- 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.

1 25

[Bug 38530] New: unimplemented function msvcp100.dll.?_Mtx_new@threads@stdext@@YAXAAPAX@Z
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

https://bugs.winehq.org/show_bug.cgi?id=38530 Bug ID: 38530 Summary: unimplemented function msvcp100.dll.?_Mtx_new(a)threads@stdext@@YAXAAPAX(a)Z Product: Wine Version: 1.7.42 Hardware: x86 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown Assignee: wine-bugs(a)winehq.org Reporter: xqwerty123(a)luukku.com Distribution: --- Direct download. http://download.esignal.com/products/esignal/eSignal_12.1.3756_x86.exe You only need to install it, all default. -- 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.

1 4

[Bug 28193] New: wineconsole does not support RTL locales
by wine-bugs＠winehq.org Oct. 14, 2016

Oct. 14, 2016

http://bugs.winehq.org/show_bug.cgi?id=28193 Summary: wineconsole does not support RTL locales Product: Wine Version: 1.3.26 Platform: All OS/Version: All Status: NEW Severity: normal Priority: P2 Component: programs AssignedTo: wine-bugs(a)winehq.org ReportedBy: fgouget(a)codeweavers.com I am entering this bug on behalf of Yaron Shahrabani because I did not find any other bug describing this issue. So the issue is that wineconsole does not support the RTL languages such as Hebrew and Arabic. To test this do the following: * xcopy is a tool that is Unicodified and properly accesses the console. So we are going to use it for testing. * Make sure your system is configured with support for the Hebrew locale ('dpkg-reconfigure locales' on Debian). * Hack po/he.po to add a Hebrew string with a trailing '.\n' for the xcopy usage message (it does not matter what it says). * Recompile xcopy. * First check that 'xcopy /?' shows the right string, very likely with the wrong orientation, in the terminal emulator: LANG=he_IL.utf8 ./wine programs/xcopy/xcopy.exe.so /? If it still shows the English usage it may be because you are missing libgettextpo. Other display issues may be because you are missing the right fonts or because your terminal emulator is buggy (Gnome terminal >= 3.0.1 and xterm >= 271 should show the string at least). Remember that the Unix terminal emulators are known to have a buggy RTL implementation and thus will display the message with the wrong orientation. * Then check that the right string shows up in wineconsole: LANG=he_IL.utf8 ./wine wineconsole cmd then type xcopy /? In my test the line where the Hebrew string should appear is almost empty (I only see the '32' and trailing '.' that were part of the translated string). It's possible there are two problems: * A font selection issue which could explain the 'invisible' characters. * An LTR vs. RTL issue: the '.' at the end of the PO string appears at the end of the wineconsole line which seems wrong since the same dot appears at the start of the line in a MessageBox(). The same test can be done with an Arabic locale by editing the po/ar.po file and setting LANG=ar_MA.utf8 for instance. As a last data point, a similar test with the left-to-right Telugu locale (LANG=te_IN.utf8, po/te.po) gets me squares in the Gnome terminal (proving my system has a font issue for that locale), but also in wineconsole. So unlike for Hebrew I did not get 'invisible characters'. -- 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.

1 20