Re: [E-devel] two small imlib2 bugfixes

[Piotr <pp...@si...>]

On Wednesday 06 November 2002 05:15, Michael Jennings wrote:
> > The first patch is self-evident, there is a fix for realloc() of
> That was fixed ages ago.

Oh, now I see it, 6 months ago!! I am terribly sorry, I had probably=20
looked into wrong directory :((

> > The second patch needs more explanation. I have removed these two
> > fields from tga_footer struct because they:
> Sorry, this breaks TGA loading on my system.  The existing loader
> works fine.

But here I am puzzled. I use gcc 2.95 to compile this loader. If=20
compiled without optimizations it works fine. When I turn on=20
optimizations here is what happens: gcc adds padding bytes at the end=20
of the tga_footer

typedef struct {=20
   unsigned int extensionAreaOffset;=20
   unsigned int developerDirectoryOffset;=20
   char signature[16];=20
   char dot;=20
   char null;=20
// here padding bytes are added, so when you place another struct=20
after this one ints are nicely aligned
} tga_footer;=20

Because size of the struct changes, wrong part of the file is read=20
into tga_footer struct:

/* read the footer first */
fseek (fp, 0L - (sizeof (tga_footer)), SEEK_END);
if (fread (&footer, sizeof (tga_footer), 1, fp) !=3D 1)

sizeof(tga_footer) is bigger than the real footer size in file, so it=20
seeks too far and wrong part of the file is read.

I have removed those ints in hope that gcc in that case won't add any=20
padding bytes. It works for me. But now I am really surprised why it=20
doesn't work for you. Does gcc add padding bytes anyway? Are there=20
also some padding issues in the TGA format? Can you send me those=20
images you were testing the loader on?

Cheers,
--=20
 Piotr Paw=B3ow
 mailto:pp...@si...