`l3opacity`: `\opacity_stroke:n` and `\opactiy_fill:n` not working and/or leading to broken PDF about latex3

[muzimuzhi]

The problem of broken PDF can be fixed by initializing two internal token variables (in l3backend-opacity.dtx or as a workaround after backend driver is loaded).

\tl_set:Nn \l__opacity_backend_fill_tl { 1 }
\tl_set:Nn \l__opacity_backend_stroke_tl { 1 }

[muzimuzhi]

@josephwright Currently m3opacity001.tlg is partially broken since a \DocumentMetadata{} is missing in m3opacity001.lvt.

latex3/l3experimental/l3opacity/testfiles/m3opacity001.tlg

Lines 17 to 19 in b492b56

	! Undefined control sequence.
	<argument> ..._stroke_tl {}\pdfmanagement_add:nnn
	{Page/Resources/ExtGState}...

Have a word about \DocumentMetadata{} in l3opacity's doc?

[josephwright]

@muzimuzhi Was just working on something else, but I had a feeling where was a missing init step - you want to fix?

[jasperhabicht]

\tl_set:Nn \l__opacity_backend_fill_tl { 1 }
\tl_set:Nn \l__opacity_backend_stroke_tl { 1 }

Nice, that indeed solves the problem of the broken PDF. But in MWE B the stroke is still not transparent, if I am not mistaken.

[muzimuzhi]

But in MWE B the stroke is still not transparent, if I am not mistaken.

No, some use case is still not working. I'm digging.

[josephwright]

Looking back, I'm wondering why I use a tl here - I'll probably change ;)

[muzimuzhi]

Ah ha, a typo:

--- a/l3backend/l3backend-opacity.dtx
+++ b/l3backend/l3backend-opacity.dtx
@@ -237,7 +237,7 @@
           { opacity.fill #1 }
           { << /ca ~ #1 >> }
         \pdfmanagement_add:nnn { Page / Resources / ExtGState }
-          { opacity.stroke #1 }
+          { opacity.stroke #2 }
           { << /CA ~ #2 >> }
 %<*dvipdfmx|xetex>
         \__kernel_backend_literal_pdf:n

[josephwright]

@muzimuzhi You want to fix in a PR or should I just do it directly?

[muzimuzhi]

Looking back, I'm wondering why I use a tl here - I'll probably change ;)

Because str functions are slower? (I didn't check l3color.) Or perhaps you want to use fp instead.

@muzimuzhi You want to fix in a PR or should I just do it directly?

Considering the possible re-typing, you can do it directly.

[josephwright]

Looking back, I'm wondering why I use a tl here - I'll probably change ;)

Because str functions are slower? (I didn't check l3color.) Or perhaps you want to use fp instead.

Should really be an fp, yes - but as we don't do maths, I see why I did it this way now. I'll fix shortly.