There are useful test file which we can't add to comparefiles since they didn't embed CID font. The test file of the bug 687970 is an example.
I meant "there are useful files". Setting P3 since it impedes the development.
As the regression testing is being modified to simplify the script, the ability to support this feature will be considered. The desire is to avoid the need for too much modification of the ghostscript under test (such as a cidfmap that doesn't match that in svn). My initial concept is to add some tests that *include* the installation of a CIDFont (such as WadaGo-Bold) as a prefix to a file that uses _that_ font without the need for cidfmap. We can modify the test file so that WadaGo-Bold is used directly (e.g., modify the file of bug 687970). Note that we should also 'fix' the KozMin-Bold_NonEmbed.pdf file to make it useful by doing a installing a CIDFont prior to the actual PDF, possibly changing the PDF to use that font. BTW, leaving priority at P3, but specific need to be provided to describe exactly how this 'impedes development' since I expect developers to be able to set up special tests in a few seconds when working in a particular area. The point of regression tests is _not_ to automate all testing, but to spot unforeseen side effects.
When I wrote "impedes development", I meant that currently files that use a CID font and don't embed it can't be added to comparefiles. So code changes that affect such files may cause an undetected regression. I don't mean to test a dynamic CID font installation. I just want to have some predefined fonts in regression/gs/Resource/CIDFont . After that we can easy add documents which use those fonts. Another interesting question is how to test files, which need a special font substitution. One way is to create a "standard" set of CID fonts and install it permanently to the testing environment (as we do with fonts), and update gs/lib/cidfmap when a document is added to comparefiles. It means that the testing framework will use a diffgerent cidfmap than the distribution. A simpler way is to include the "standard" set into distribution - I recall it was discused many times with no real progress.
Ray please make it go. We're loosing lots of man*hours due to no regression testing of installed CID font and CID font emulation with True Types. All we need here is to insert copying few additional resources into regression scripts (maybe scp from casper).
Added Henry to CC list. Henry please pay attention to this issuew because iot is important for me Alex and Ken. IMO we need at least 2 fonts : a FIDFontType 9 one and a True Type.
Please specify the desired fonts and associated cidfmap. We have the CIDFonts at http://www.artifex.com/AsianFonts or we could use some TTF fonts from Windows. Also Artifex has 1 licensed copy of HeiseiKakuGo-Bold (or some name similar to that from Adobe). This one, iirc, is a OTF font.
For the beginning I would be satisfied if we have any 1 CIDFontType 1 and any 1 TrueType font. It may be improved later on necessity. More interesting question is how to supply a specific cidfmap for a particular test. I'll think what to mchage in Init to allow to include one into a PS document.
Someday. If the submitter feels this is critical, please feel free to add CID fonts to the regression testing.
Dumping this on to Marcos. If advice is needed, please consult Masaki or Ken as to what should be tested (or close as WONTFIX).
Chris suggests that since we no have cid font fallback that adding file(s) with non-embedded cid fonts would satisfy this bug. I agree and will and some appropriate test files.