The difference has been validated with diffpdf.
The following could not be tested and were left unchanged:
AEA.lyx
IJMPC.lyx
ja_beamer-conference-ornate-20min.lyx
The following is the script that I used (in lib/templates):
LYX=../../build/src/lyx
$LYX -E pdf2 $1.old.pdf $1
sed -i "s/^\\\\begin_inset Separator parbreak$/\\\\begin_inset Separator plain/" $1
sed -i "/^\\\\begin_inset Separator latexpar$/ { N; d; }" $1
$LYX -e lyx $1
$LYX -E pdf2 $1.pdf $1
diffpdf $1.old.pdf $1.pdf
Some templates had the inputenc setting "default" (meaning
use the language default encoding, (not TeX's default which is 7-bit ASCII) but
don't load inputenc, leave handling of non-ASCII characters to the author).
This is no apparent problem as long as no non-ASCII character is used in the
document. It is very unsave, however, as any non-ASCII character used in the
document has the potential to either make compilation fail or be
misrepresented in the output.
Also MacOSX ReadMe files. Note that a few lib/doc files are
also "updated" because trailing spaces are removed, but their file
formats are the same because they were recently updated at 83672113.
I did "git checkout LFUNs.lyx" because this file is generated
automatically and has a special header.
Manuals, examples and templates that use (traditional) bibtex should have set the bibliography processor to bibtex. Else, these documents do not compile if a user has set the processor to 'biber' in the preferences
