Open paulrouget opened 2 years ago
I have really looked at how the mupdf integration works,
You meant you haven't?
but I was wondering if it would make sense to use https://github.com/messense/mupdf-rs instead of relying on the locally built mupdf library?
I don't think it'll work:
/usr/local
.I have really looked at how the mupdf integration works,
You meant you haven't?
Right… typo.
I don't think it'll work
Understood. Thanks.
I'm still trying to wrap my head around this.
I'm patching the source.
Can you explain a bit what this is about? What's the evenodd
change for?
I need to have precise control over the size
What about linking statically? All the unused symbols would be dropped.
I'm still trying to wrap my head around this.
I'm patching the source.
Can you explain a bit what this is about? What's the
evenodd
change for?
It's a commit from upstream that came after the 1.20.0 release. It fixes the handling of the fill-rule
attribute in SVG paths. But I guess I could wait until the next release since I'm not making new icons very often.
I need to have precise control over the size
What about linking statically? All the unused symbols would be dropped.
Since I'm referencing fonts stored inside the library: won't the plato binary size increase significantly?
Since I'm referencing fonts stored inside the library: won't the plato binary size increase significantly?
Oh then I misunderstand the point of the wrapper :)
Since I'm referencing fonts stored inside the library: won't the plato binary size increase significantly?
plato statically linked with mupdf should be smaller than plato + mupdf.so. So the overall footprint should be smaller?
I'm just curious :) Thank you.
Since I'm referencing fonts stored inside the library: won't the plato binary size increase significantly?
Oh then I misunderstand the point of the wrapper :)
- mupdf is shipped along with plato, right?
Yes.
It's not builtin in Kobo system?
No.
- plato is linked to mupdf. What does the wrapper does then?
MuPDF uses C intrinsics (setjmp
and longjmp
) to implement its error handling mechanism, since we can't deal with this within Rust, we have to write C wrappers for all the functions that might fail (mostly all the functions that take an fz_context pointer as argument according to the documentation) and bind to the wrapper instead of the wrapped. mupdf-rs has its own wrapper.
Since I'm referencing fonts stored inside the library: won't the plato binary size increase significantly?
plato statically linked with mupdf should be smaller than plato + mupdf.so. So the overall footprint should be smaller?
I trust you on that, but having a smaller plato binary is more important to me.
Understood. Thank you.
I have really looked at how the mupdf integration works, but I was wondering if it would make sense to use https://github.com/messense/mupdf-rs instead of relying on the locally built mupdf library?