[scribus] Odd glyph issue when exporting to PDF

[John Jason Jordan]

I have been trying to assist someone on an OpenOffice.org e-list. He is
finding missing glyphs in his favorite font. Now, I know OOo will fake
glyphs by borrowing a missing glyph from another font. Why and how it
does this has always been a mystery to me, but it does do it and it is
a frequent source of problems. Scribus, on the other hand will reliably
display only what is actually in the font, at least in canvas view.

The individual included some glyphs in an e-mail and I copied and
pasted them into a new OOo document. I did so five times and applied
the individual's problem font to one line, and four of my other similar
fonts to the other lines. All the glyphs appeared. Then I pasted from
OOo into a new document in Scribus 1.3.5.1 (on Fedora x86_64). As
I expected, two of the glyphs appeared in two of the fonts, but all the
rest appeared as empty boxes.

I was intending to respond to the individual by e-mailing him PDFs from
both OOo and Scribus to show him how unreliable OOo is when it comes to
fonts. But when I exported as PDF from Scribus I got very strange
results.

First, Preflight Viewer complained that there were missing glyphs. I
told it to ignore the problem and proceeded to export the PDF. When I
opened the PDF in Adobe Reader 9.1 it gave me a warning that there were
problems with the PDF. But it opened the document anyway. However, for
each of the missing glyphs Scribus had apparently inserted a capital
letter of the font name. That is, it did so for all but one of the
fonts. The four fonts it did so are all TrueType fonts, and one it did
not do so for is OpenType. 

I got the same display in Okular, Foxit Reader, Evince and Cabaret,
although only Adobe Reader gave me a warning about errors in the file.

The first time I tried it the Scribus export to PDF utility embedded
three of the fonts and converted two to outlines. I didn't bother to
fix the embedding rules, I just exported again and told it to convert
all to outlines. Either way I got the same results.

I'll copy and paste here, although I suspect the internet Gods will
have a seizure and it will come in messed up in your mail readers.
First, here is from test.sla:

𝄆    𝄇   ♯   ♭   𝄁   𝄎   𝄂   Junicode
𝄆    𝄇   ♯   ♭   𝄁   𝄎   𝄂   FreeSans
𝄆    𝄇   ♯   ♭   𝄁   𝄎   𝄂   Bitstream Vera Sans
𝄆    𝄇   ♯   ♭   𝄁   𝄎   𝄂   Arial Unicode MS
𝄆    𝄇   ♯   ♭   𝄁   𝄎   𝄂   Liberation Sans

And here is copied and pasted from the PDF, first with Evince and then
with Adobe Reader:

J J J J J  J   J J  J J  J J   Junicode
FF  FF  ♯ ♭ FF     FF   FF   FreeSans
BB   BB  B   B   BB   BB   BB    Bitstream Vera Sans
L L L L L  L   L L  L L  L L   Liberation Sans

J J J J J J J J J J J J Junicode
F F F F   F F F F F F FreeSans
B B B B B B B B B B B B Bi tstream Vera Sans
L L L L L L L L L L L L Liberation Sans

Oddly, none of my PDF viewers allow me to copy and paste the Arial
Unicode MS line. That is the only font that is OpenType as well.

I cannot check the fonts with FontForge or Fontmatrix because I don't
know the Unicode points for the glyphs in question.

The big question is why does Scribus insert the uppercase letter of the
font name for each missing glyph when exporting as PDF. That is just
wrong. If the glyph ain't there, it ain't there. And a sub-question is
why does it do it properly for OpenType fonts?

It may be a bug in Scribus, but I thought I'd ask here first.