Re: WOFF2 failure

Sorry, that's my fault. I didn't want to remove the old API as doing so
would break anyone using it and the repo is public so it's not easy to do a
sweeping update of all consumers.

As Khaled suggests, please use this (from
https://github.com/google/woff2/blob/master/src/woff2_dec.h):

// Decompresses the font into out. Returns true on success.
// Works even if WOFF2Header totalSfntSize is wrong.
bool ConvertWOFF2ToTTF(const uint8_t *data, size_t length,
                       WOFF2Out* out);

And avoid this; it's more fragile:

// Decompresses the font into the target buffer. The result_length should
// be the same as determined by ComputeFinalSize(). Returns true on
successful
// decompression.
bool ConvertWOFF2ToTTF(uint8_t *result, size_t result_length,
                       const uint8_t *data, size_t length);

Rod S

On Wed, Sep 21, 2016 at 12:20 PM, Khaled Hosny <khaledhosny@eglug.org>
wrote:

> LOL, I actually first wrote a much detailed answer then thought it is
> too newbish and discarded it.
>
> Yes, you will need to use the new API that uses the WOFF2Out object as
> it does not suffer from the limitations of the old API WRT trusting
> totalSfntSize and origLength.
>
> Regards,
> Khaled
>
> On Wed, Sep 21, 2016 at 11:31:24AM -0700, Ned Holbrook wrote:
> > Thanks, but you’ll have to be a bit more explicit for us poor newcomers.
> :) The fix in this case would be to use the ConvertWOFF2ToTTF() overload
> taking a WOFF2Out object?
> >
> > > On Sep 21, 2016, at 11:25 AM, Khaled Hosny <khaledhosny@eglug.org>
> wrote:
> > >
> > > On Wed, Sep 21, 2016 at 11:11:12AM -0700, Ned Holbrook wrote:
> > >> This is regarding <http://test.chrislewis.codes/woff2/>, found on <
> http://typedrawers.com/discussion/1775/browser-ots-
> rejections-for-woff2-but-not-woff>; the WOFF2 for the second line of the
> test page is not being used.
> > >>
> > >> I stepped through I discovered the failure is due to this code in
> woff2_dec.cc:
> > >>
> > >>    if (PREDICT_FALSE(static_cast<uint64_t>(table.dst_offset +
> table.dst_length)
> > >>> out->Size())) {
> > >>      return FONT_COMPRESSION_FAILURE();
> > >>    }
> > >>
> > >> At the point of failure:
> > >>    (lldb) p out->Size()
> > >>    (size_t) $7 = 405520
> > >>    (lldb) p table.dst_offset + table.dst_length
> > >>    (unsigned int) $8 = 417149
> > >>
> > >> Now, Khaled commented on the TypeDrawers page to suggest this check is
> > >> overly strict. Does this mean the reference implementation needs to be
> > >> fixed?
> > >
> > > See:
> > > https://github.com/google/woff2/pull/48
> > >
> > > Regards,
> > > Khaled
>
>