Age | Commit message (Collapse) | Author |
|
Correct a regression introduced in 7557f17ed2 which caused the Dreamcast
to hang at the license screen when running SCUMMVM.BIN from a CD.
7557f17ed2 fixed configure's feature detection by moving crt0.o from
LDFLAGS to LIBS, but that changed the linking order resulting in the
start symbol from crt0 being located at an address other than
0x8c010000. ELF binaries loaded over serial or TCP/IP were unaffected
(presumably because the loaders use the start address from the ELF
header instead of a fixed address) which is probably why this went
unnoticed.
This commit corrects the link order by moving crt0 back into LDFLAGS and
adds "-lronin -lm" to LDFLAGS to ensure that configure checks continue
to work.
|
|
Remove $LDFLAGS and $CXXFLAGS from cc_check call in translation support
check. Both variables are redundant because they are already expanded
in cc_check, and expanding LDFLAGS twice causes duplicate symbol errors
for targets that include crt0.o in LDFLAGS.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This should improve linker time as well as reduce the size of the
built binaries (though this does not affect runtime memory usage
since the debug segments are not loaded into memory except by a
debugger) whilst still giving debug information.
|
|
$(ronindir) and the explicit crt0 may have happened to work because
library detection state for libmad was forced, but it was always
breaking the feature detection of the build system. Now we can
compile to Dreamcast using the normal detection system.
|
|
|
|
|
|
|
|
|
|
As of freetype-2.9.1 the freetype-config script no longer gets installed
by default.
|
|
|
|
Previously, only CXX could be overridden by the environment, which
made it impossible to correctly set tools for cross-compiler
toolchains which were not compatible with the default system tools.
|
|
|
|
|
|
|
|
macOS' clang does not support this directive (yet), which
leads to tons of useless warnings
|
|
Clang 6 enables -Wpragma-pack which warns when leaving an included file
which changes the current alignment.
Our common/pack-{start,end}.h trigger this and it cannot easily and portably
be disabled inside pack-{start,end}.h so we disable it globally for now.
|
|
|
|
|
|
|
|
|
|
|
|
The OPL2LPT is an OPL2 chip plugged on a parallel port. It is
write-only but otherwise acts as a classic AdLib. This commit adds
support for this device.
User is expected to have the right permissions on the parallel
port. By default, the first suitable parallel port is used. It is
possible to override that with the hidden configuration setting
"opl2lpt_parport".
It depends on the presence of the libieee1284 library which abstracts
a bit parallel port handling. An alternative would be to access
directly /dev/parportX on Linux. This would amount of code but it
would be Linux-only.
Tested with Indy 3 and SOMI.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Configure will still build every target with shared objects by default.
Done:
- Added a configure switch (--enable-static) for static builds
- Right now only tested for SDL targets (checked working on AmigaOS4 with SDL1/2)
Missing:
- Add the platforms that need the switch for cross-compiling (macOS, iOS, probably more)
(I can´t do that myself as i have never cross-compiled and wouldn´t be able to check)
@csnover asked on the forums to create a PR for it, so, with my limited (non-existant is more like it) knowledge, i wrapped something up.
I *know* that it isn´t enough, as there already is some kind of static builds flag in configure (_engines_built_static). As i don´treally know if this is the same or a different cause (static plugins?), i´m not touching it.
But please feel free to add or change what needs to be done.
|
|
Either the newer munt, the newer toolchain or the better compiler flags
made it run well enough.
|
|
-mno-fp-in-toc is enough not to overflow the TOC. According to the GCC
documentation this allows for faster code to be generated than with
-mminimal-toc.
|
|
|
|
|
|
PORTS: Remove the Zaurus and Yopy ports
|
|
|
|
No official release was ever done for that platform. Production ceased
in 2005.
|
|
|
|
|
|
|
|
|
|
This reverts commit 2a22c16b916602ee9ced429455add5707b0c4f74.
The old Buildbot (or maybe the build code's messing about with
LDFLAGS) seems to be broken and won't link successfully with the
std APIs, so this gets turned back off again until the new Buildbot
is active during the next release cycle.
|
|
Modern Android CPUs are more than powerful enough to emulate the
MT-32, and users are requesting its support.
|
|
Refs Trac#6132, Trac#6810, Trac#6957. Closes gh-1067.
|