The disassembler was producing:
strh r4, [r0, #2]
instead of
strh r4, [r0], #2
which was causing the alignment and index to go all wrong causing crashes.
Tested-by: dougt
Adrian Johnson (3):
Fix win32 build breakage
Ensure win32 font index_to_ucs4() sets ucs4 to -1 if lookup fails
Correct the comment for the index_to_ucs4 font backend function
Behdad Esfahbod (2):
Err, make gtk-doc happy again
[user-font] Add CAIRO_STATUS_USER_FONT_NOT_IMPLEMENTED
Chris Wilson (37):
[build] Enable shave support
[build] Add lcov-1.7 to known list
[image] Return FORMAT_INVALID for an error surface.
[image] Make _cairo_image_analayze_transparency() more format agnostic
[image] Treat A1 image surfaces as BILEVEL_ALPHA
[test] Create a new fallback surface to exercise 16-bit paths.
[png] Coerce FORMAT_INVALID to a known image format
[surface] Convert FORMAT_INVALID during snapshots
[xlib] Use a similar content surface for cloning images
[build] Report status of gtk-doc in summary
[build] Fix gtk-doc interoperation with shave
[check] Quieten output during checking headers-standalone
[cairo] Fix errors reported by check-doc
[cairo] Convert C99 comments to /* ... */
[cairo] Remove stray semi-colon
[cairo] Protect typeof against -ansi
[skiplist] Missing include for ffs()
[script] Missing include for INT32_MAX
[script] Pedantic violations
[path] Define _BSD_SOURCE to enable hypot()
[doc] Update identifiers
[ft] Restore the ability to lazily resolve patterns.
[xlib] Use minimal depth for similar clones.
[test] Fix typos that excluded backend test sources
[test] Add a note to "Getting the elusive zero failures"
[test] Relax APPROX_EQUALS condition
[test] Check text->path with user-fonts (twin)
[test] Clear expected floating point exceptions
[test] Stress the ft from-face cache
[memfault] Update macros to avoid namescape collision with memcheck
[debug] Check image contents using memcheck.
[win32] Typo
[win32] Wrap win32-font usage with CAIRO_HAS_WIN32_FONT
[surface] Provide nil-surface for INVALID_SIZE
[surface] Early return if the backend cannot clone an image
[xlib] Handle too-large images gracefully.
[in-fill] Treat on-edge queries as inside.
Jeff Muizelaar (1):
Only include <strings.h> for ffs() when we HAVE_FFS
Karl Tomlinson (1):
[ft] Resolve mutual referencing problems with zombie faces
cairo_scaled_font_init() allows rank 0 matrics so cairo_set_font_matrix
should too. This keeps us from accidentally breaking our cairo context
and causing things to draw wrong.
