Fix signature of VCustomSource#SeekCallback (fixes #171) #172
Conversation
jbaiter
force-pushed
the
fix-custom-source-seek
branch
from
3c7f948 to
41e80dc
Compare
Previously the offset was mistyped as a `*void`/`MemorySegment`. It was now changed to be a `gint64`/`long`. Additionally, the `int whence` parameter was replaced with a new `SeekWhence` enum that provides more type safety and discoverability for implementers. To test, a sample was added to provide a `VCustomSource` based on a `FileHandle` that implements seeking, testing it against a TIF image.
jbaiter
force-pushed
the
fix-custom-source-seek
branch
from
41e80dc to
6eab6da
Compare
|
Looks great, thanks @jbaiter! Could you also run |
|
Done, sorry that the diff got so unwieldy, but the new enum changed the odd/even CSS classes on quite a bunch of entries 😅 I also had to heavily modify (not included in this PR) the various scripts in the repository to be able to generate the bindings on my Linux/x86-64 machine. Would you be open to a PR that updated the scripts to be cross-platform? Or maybe even run stuff like jextract in a Docker container to have a more controlled/portable environment? |
|
Yes, definitely open to cross-platform improvements in the generation scripts! I like the idea of doing it with a Docker container - I think it would be good to standardise the generation based on something known/stable, like a Debian image. No worries about the diff being larger, the doc generation always ends up like that 👍 |
|
@jbaiter This is now published as |
2 participants
Previously the offset was mistyped as a
*void/MemorySegment. It was now changed to be agint64/long. Additionally, theint whenceparameter was replaced with a newSeekWhenceenum that provides more type safety and discoverability for implementers.To test, a sample was added to provide a
VCustomSourcebased on aFileHandlethat implements seeking, testing it against a TIF image.I also tried testing it against a JP2 conversion of
fox.jpg, but for some reason that did not work with any of theVSourceimplementations, including file-based ones, so the bug is probably elsewhere (possibly in the libvips jp2k foreign loader code that tries to detect jp2 files).