implementations are provided in the same file and the no longer necessary
inlining will break the build. The compat implementations will need to be
neutered with the next bump. Also silence a harmless compiler warning
about an incorrect function pointer type. OpenSSL changed the bio_info_cb
type "for backward compatibility" while LibreSSL didn't. Instead, use the
correct BIO_info_cb.
ok sthen
mupdf now builds C++ code and requires a compiler that
supports -std=c++17, so zap COMPILER_LANGS=c.
Also zap MODGCC4_ARCHS=armv7, which doesn't do what you think it does in
this context (base-clang would still be picked up on armv7). mupdf on
armv7 should be tested again before adding another workaround.
ok kmos@ sthen@ (maintainer)
use compatible code for libressl (mupdf expects openssl 1.1.0), and use
certs from /etc/ssl/cert.pem instead of relying only on Adobe embedded
certificate.
OK sthen@
fixing build on BE arches. Tracked down by George Koehler who proposed
an alternative simpler patch (but I'd prefer this one which is closer
to the changes made for LE).
- mupdf-gl now requires building with javascript support, so is no longer
available in the default flavour (which has js disabled). If you use mupdf-gl
you will need to switch to the js flavour (pkg_delete mupdf; pkg_add mupdf--js).
- drop the patch to convert to shared libraries, the build process of
mupdf is a headache and changes rather often, so this patch is getting in
the way too much. unfortunately this bloats the packages somewhat...
libGL/libxshmfence tries to create "/tmp/shmfd-XXXXXX" files in some
cases (not seen here but reported by mikeb). just drop the pledge
from mupdf-gl completely because I can't hope to evaluate the codepaths
taken in mesa on various video hardware to come up with a reliable one.
allocators to do this locking too. However mupdf tries to manually align things
in a way that doesn't match what openjpeg does, which we noticed when frees
were segfaulting because the addresses didn't match up.
In the case of the openjpeg port it's relying on malloc(3)'s guarantee
"The allocated space is suitably aligned (after possible pointer coercion)
for storage of any type of object" so patch mupdf to do the same.
Fixes crash noticed by jca@ in https://www.broadband-forum.org/technical/download/TR-177.pdf
