Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo

Close

#7 fixes 2gb+ files issue on win32

Unstable (example)
open
Peter C
5
2005-02-10
2005-02-10
Anonymous
No

It was impossible to create PARs for files larger than 2GB
on Windows. This small patch fixes this bug.

Discussion

  • Logged In: NO

    sorry, forgot to attach .diff file... here it is:

    --- par2cmdline.h.orig 2005-02-10 13:25:22.000000000
    +0100
    +++ par2cmdline.h 2005-02-10 13:25:39.000000000
    +0100
    @@ -38,7 +38,9 @@
    #include <assert.h>

    #define snprintf _snprintf
    -#define stat _stat
    +#define stat __stat64
    +#define __stat64(a, b) _stat64(a, b)

    #define __LITTLE_ENDIAN 1234
    #define __BIG_ENDIAN 4321

     
  • Logged In: NO

    Does anyone have a compiled binary with this fix?

     
  • Logged In: YES
    user_id=189212

    It is also wrong for non-Win32. I was pretty annoyed with UnRarX wouldn't
    work in Mac OS X because of bugs in this thing. The bug is:

    #define OffsetType long
    #define MaxOffset 0x7fffffffL
    #define LengthType unsigned int
    #define MaxLength 0xffffffffUL

    Something like this should be much more useful in operating systems like
    Mac OS X that support 64-bit stat/seek/*:

    #define OffsetType off_t
    #define MaxOffset (sizeof(off_t) == 8 ? 0x7fffffffffffffffL : 0x7fffffffL)
    #define LengthType fpos_t
    #define MaxLength (sizeof(fpos_t) == 8 ? 0xffffffffffffffffL : 0xffffffffL)

    For other OSes, you need to add a test for things like fseek64 and use that if
    available.

     
  • Logged In: YES
    user_id=189212

    Oops. Sorry, those should have had a LL or ULL suffix for the 64-bit values. Oh,
    and a UL for the 0xffffffff, though it has that already. Typo on my part.

     
  • Logged In: YES
    user_id=189212

    Ouch. There are a lot more mistakes than that on the UNIX side of things. The
    code also incorrectly uses size_t which is always 32 bits on a 32-bit UNIX
    platform even if it can support 64-bit files. Odds are, everywhere size_t is used,
    it should be universally replaced with off_t throughout the entire program.

     
  • Logged In: YES
    user_id=189212

    I'm wrong. The uses of size_t are fine. I'm getting some bogus compiler
    warnings, but they do appear to be bogus.

     
  • Logged In: YES
    user_id=189212

    Well, two mistakes with this. First, need to change fseek to fseeko in all three
    places that it is used in diskfile.cpp, since fseek takes a 32-bit long.

    Second, I think I was wrong about the change to LengthType and MaxLength.
    I'm pretty sure only OffsetType and MaxOffset need to change.

    After making those fixes, I have verified that it can, in fact, fix a corrupted
    archive of well over 4 GB in size.