[libbluray-devel] [PATCH 2/4] The number of bytes used by fixed-length character fields is stored as Uint8 (ECMA 1/7.2.12).

[Petri Hintukainen]

pe, 2017-06-02 kello 08:57 +0200, Andreas Zelend kirjoitti:
> 2017-06-01 9:49 GMT+02:00 Petri Hintukainen
> <phintuka at users.sourceforge.net>:
> > ke, 2017-05-31 kello 12:33 +0200, Andreas Zelend kirjoitti:
> > > I'll prepare another patch regarding off_t size problems.
> > 
> > in libudfread ?
> > 
> > off_t is used in only one file, maybe you can just define it to
> > uint64_t or something else. Or add native Windows API versions of
> > the
> > remaining functions.
> > 
> > 
> 
> Yes, maybe the following:
> 
> From 6ba9f18fbb4cfa9994a7c33d30750d30daff2fc4 Mon Sep 17 00:00:00
> 2001
> From: ace20022 <ace20022 at ymail.com>
> Date: Thu, 14 Jan 2016 15:46:43 +0100
> Subject: [PATCH] Use 64-bit version of lseek and off_t for win32.
> 
> ---
>  src/default_blockinput.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/src/default_blockinput.c b/src/default_blockinput.c
> index 4daa6a3..aaa8c62 100644
> --- a/src/default_blockinput.c
> +++ b/src/default_blockinput.c
> @@ -41,6 +41,10 @@
>  #include <stdio.h>
>  #endif
>  #include <io.h>
> +# undef  lseek
> +# define lseek _lseeki64
> +# undef  off_t
> +# define off_t int64_t
>  #endif

looks OK.

> I've only tested it with msvc, not with mingw.

I think lseek() is incorrect in mingw too. It was unnoticed because of
it is used only with broken media (when backup descriptors are read
from the end of the disc).

> The issue that needs
> fixing is the right shift at line 67 ov.OffsetHigh = (offset >> 32); 
> which is undefined (nop in my case) if offset is a 32-bit variable.

There's something wrong with off_t definition. It simply doesn't work
if it is not 64 bits.

> > > btw do you plan to tag a release in the near future?
> >
> > libudfread ?
> > We could tag 1.0.0, there haven't been much changes recently.
> >
> > libbluray ?
> > Probably sooner than later, there's already a pile of fixes since
> > previous release.
>
> I've found a crash yesterday when calling bd_get_event between
> bd_init and bd_open* , since bd->disc is null. Are you interested?

I'll fix it. It kind of makes sense to init event handling before open.

> Having a new release for both libs would be great.

Yes. I need to stop creating new bugs :)