On 4/6/2013 10:55 AM, Patrick Gundlach wrote:

...

...

anybody got a 64 bit binary for windows (7) 64? I've got some problems with the current 32 bit binary and I was wondering if using a 64 bit binary would fix them.

I've tried some time ago with no luck, but I would like to spend more time on this. Which kind of problems have you seen ?

the same as on Januar 16 on this list:

!LuaTeX error (file C:\Users\.../share/fonts/texgyreheros/texgyreheros-regular.otf): Parsing CFF DICT failed. (error=-1) ==> Fatal error occurred, no output PDF file produced!

all other platforms are OK, win32 on 32bit is also OK.

I can't produce a MWE at the moment.

I also sometimes get that error (also with that font). But we need a minimal example as then Taco can run detailed mem tests. It could as well be that on other platforms the same issue is obscured (we had similar issues, unitialized variables or so that then depend on other conditions.) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | voip: 087 875 68 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------

...

I also sometimes get that error (also with that font).

I've tried with different fonts (lin libertine, arial.ttf), I always get the same error. My self-compiled binary (32bit) tells me something like: "premature end of file".

forget that. Using Akira's (?) binaries are better... Patrick

Am 06.04.2013 um 23:31 schrieb Hans Hagen :

On 4/6/2013 4:30 PM, Patrick Gundlach wrote:

...

...

...

I also sometimes get that error (also with that font).

I've tried with different fonts (lin libertine, arial.ttf), I always get the same error. My self-compiled binary (32bit) tells me something like: "premature end of file".

forget that. Using Akira's (?) binaries are better...

indeed Akira's ...

but I had that same error a while ago, so there is definitely an issue someplace

(still working on a minimal example) I get the error when using a ttf (like arial.ttf) together with a .otf or .pfb file with LuaTeX. Using ttf alone is OK, using .otf/.pfb without using .ttf seems to be OK as well. Patrick

Am 06.04.2013 um 10:55 schrieb Patrick Gundlach :

!LuaTeX error (file C:\Users\.../share/fonts/texgyreheros/texgyreheros-regular.otf): Parsing CFF DICT failed. (error=-1) ==> Fatal error occurred, no output PDF file produced!

I've been able to reproduce the situation on my Mac with a pretty-much-minimal-testfile. http://tracker.luatex.org/view.php?id=820 Patrick

...

I've been able to reproduce the situation on my Mac with a pretty-much-minimal-testfile.

http://tracker.luatex.org/view.php?id=820

Wfm (10.8.3), sorry.

that's great! ... but I now I get the next error (perhaps related) on Windows 7 64 bit (report submitted). We should not get bored ;-) Thanks for the great support so far! Patrick

Taco Hoekwater writes:

Hi,

On Apr 8, 2013, at 11:59 AM, Patrick Gundlach wrote:

...

...

...

I've been able to reproduce the situation on my Mac with a pretty-much-minimal-testfile.

http://tracker.luatex.org/view.php?id=820

Wfm (10.8.3), sorry.

that's great! ... but I now I get the next error (perhaps related) on Windows 7 64 bit (report submitted).

We should not get bored ;-)

I do not have any win7 machine any more (my wife now has Win8, and pretty much nothing works 'as normal' on that ..), which makes debugging pretty much impossible for me.

I'm keeping my significant other operating system user on Windows NT. -- David Kastrup

On Wed, Apr 10, 2013 at 11:16 AM, Fabrice Popineau wrote:

I compiled one too with MingW64 and gcc-4.8.0 Only one glitch during the compilation (which lasts forever !) : there is a clash between eof() declared in texk/web/lib/lib.h and some other eof() in system header files. I added :

#define eof weof

in texk/web/lib/lib.h and that is all.

I compiled a huge pile of (LaTeX, Tikz, OpenType Cambria, CambriaMath, Consolas fonts) slides with the resulting luatex.exe and didn't noticed any problems up to now.

Fabrice Yes, I also noticed this but I patched the io.h system header of mingw64 with #ifndef NO_EOL_OLDNAME int __cdecl eof(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; #endif

Also my ld doesn't like -Wl,--large-address-aware I have gcc version 4.6.3 (GCC) under ubuntu: are you using mingw64 under MS ? Which exception handling have you used ? (see http://qt-project.org/wiki/MinGW-64-bit) I don't know if eof is used in TeXLive . There is an eof also in common.defines, is it used by web2c ? I don't know, so I've used a dedicated mingw64 toolchain only for luatex.exe 64bit. -- luigi

On Wed, Apr 10, 2013 at 11:40 AM, Fabrice Popineau wrote:

...

Also my ld doesn't like -Wl,--large-address-aware

I have no such option in my source tree ?

Ah wait, you are right. Under linux I've modified the build.sh of luatex (from svn) if [ "$MINGWCROSS" = "TRUE" ] then B=build-windows LUATEXEXE=luatex.exe OLDPATH=$PATH PATH=/usr/mingw32/bin:$PATH CFLAGS="-mtune=nocona -g -O3 -DNO_EOL_OLDNAME $CFLAGS" CXXFLAGS="-mtune=nocona -g -O3 $CXXFLAGS" : ${CONFHOST:=--host=x86_64-w64-mingw32} : ${CONFBUILD:=--build=i686-linux-gnu} STRIP="${CONFHOST#--host=}-strip" #LDFLAGS="-Wl,--large-address-aware $CFLAGS" LDFLAGS="$CFLAGS" export CFLAGS CXXFLAGS LDFLAGS fi and I call it with bash build.sh --mingw

...

I have gcc version 4.6.3 (GCC) under ubuntu: are you using mingw64 under MS ?

Yes, I'm using MingW64 under Windows (8 !)

...

Which exception handling have you used ? (see http://qt-project.org/wiki/MinGW-64-bit)

SEH, Rubenb build.

good, it should be ok also for luajittex. -- luigi

On Wed, Apr 10, 2013 at 11:51 AM, Fabrice Popineau wrote:

I called the build script this way :

build.sh --host=x86_64-w64-mingw32 --build=x86_64-w64-mingw32

(no --mingw option). ok, in this case you should have a build directory, while with --mingw you have a build-windows directory. Not a big issue, the problem is how to manage eof: need Taco here. Can you put a note in http://tracker.luatex.org ? -- luigi

On Wed, Apr 10, 2013 at 12:09 PM, Taco Hoekwater wrote:

Well, it is definitely better to change the texmf/lib/lib.h file than to change the mingw header io.h, Ah yes, but my pov is that here luatex is right and mingw is wrong :-)

but is there no way to prevent io.h from defining eof() via gcc the command line? out of jokes: there is a NO_OLDNAMES (NO_EOL_OLDNAME is mine)

#ifndef NO_OLDNAMES int __cdecl access(const char *_Filename,int _AccessMode) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl chmod(const char *_Filename,int _AccessMode) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl chsize(int _FileHandle,long _Size) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl close(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl creat(const char *_Filename,int _PermissionMode) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl dup(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl dup2(int _FileHandleSrc,int _FileHandleDst) __MINGW_ATTRIB_DEPRECATED_MSVC2005; #ifndef NO_EOL_OLDNAME int __cdecl eof(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; #endif long __cdecl filelength(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl isatty(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl locking(int _FileHandle,int _LockMode,long _NumOfBytes) __MINGW_ATTRIB_DEPRECATED_MSVC2005; long __cdecl lseek(int _FileHandle,long _Offset,int _Origin) __MINGW_ATTRIB_DEPRECATED_MSVC2005; char *__cdecl mktemp(char *_TemplateName) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl open(const char *_Filename,int _OpenFlag,...) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl read(int _FileHandle,void *_DstBuf,unsigned int _MaxCharCount) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl setmode(int _FileHandle,int _Mode) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl sopen(const char *_Filename,int _OpenFlag,int _ShareFlag,...) __MINGW_ATTRIB_DEPRECATED_MSVC2005; long __cdecl tell(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl umask(int _Mode) __MINGW_ATTRIB_DEPRECATED_MSVC2005; int __cdecl write(int _Filehandle,const void *_Buf,unsigned int _MaxCharCount) __MINGW_ATTRIB_DEPRECATED_MSVC2005; #endif but using -DNO_OLDNAMES creates several other problems -- luigi

On Wed, Apr 10, 2013 at 12:09 PM, Taco Hoekwater wrote:

On Apr 10, 2013, at 12:00 PM, luigi scarso wrote:

...

On Wed, Apr 10, 2013 at 11:51 AM, Fabrice Popineau wrote:

...

I called the build script this way :

build.sh --host=x86_64-w64-mingw32 --build=x86_64-w64-mingw32

(no --mingw option). ok, in this case you should have a build directory, while with --mingw you have a build-windows directory. Not a big issue, the problem is how to manage eof: need Taco here.

Well, it is definitely better to change the texmf/lib/lib.h file than to change the mingw header io.h, but is there no way to prevent io.h from defining eof() via gcc the command line?

Best wishes, Taco

A patch of texmf/lib/lib.h could be /* avoid conflicts with x86_64 mingw */ #ifndef __MINGW64__ 1 extern boolean eof (FILE *); #endif -- luigi