There are two small outstanding issues with our cram test suite, which lead them to produce output whose diffs are not human-readable. We would like to tweak them a little so
that:
inconsequential differences are ignored, and thus do not visually clutter the cram test diff
we have a workable solution for gensyms that our conversion tools introduce
we have a workable solution for conversion tools that output binary formats
Two quick tweaks to our cram test suite should get us there.
Fix issues with PDF cram tests
It has recently come to our attention that our converted PDF-As, regardless of whether we use LibreOffice or pdf2archive, contain gensyms that make the data in each PDF-A
unique.
To get around this, we propose updating the cram tests for the PDF-A conversions to use pdf2text to produce "comparable" output that wouldn't confuse a cram test diff.
Fix issues with image cram tests
Putting binary image data directly into the .t files means that the diff will display weird, non-printable characters, which is not ideal for eyeballing purposes.
To get around that problem, cram tests for image conversions should base64 encode the image output and insert the base64 data into the .t file.
There are two small outstanding issues with our cram test suite, which lead them to produce output whose diffs are not human-readable. We would like to tweak them a little so that:
Two quick tweaks to our cram test suite should get us there.
Fix issues with PDF cram tests
It has recently come to our attention that our converted PDF-As, regardless of whether we use LibreOffice or
pdf2archive, contain gensyms that make the data in each PDF-A unique.To get around this, we propose updating the cram tests for the PDF-A conversions to use
pdf2textto produce "comparable" output that wouldn't confuse a cram test diff.Fix issues with image cram tests
Putting binary image data directly into the
.tfiles means that the diff will display weird, non-printable characters, which is not ideal for eyeballing purposes.To get around that problem, cram tests for image conversions should base64 encode the image output and insert the base64 data into the
.tfile.