Error trying to build under Fedora 31 with lin.build.sh

[viktormadarasz]

Hi

It gives me this when trying to build ...

viktormadarasz@fedoraws sdl-gui]$ ./lin_build.sh
gcc -O3 -ggdb -flto adb.post.c fpu.post.c mc68851.post.c mem.post.c via.post.c floppy.post.c core_api.post.c cpu.post.c dis.post.c ../core/SoftFloat/softfloat.c ../core/atrap_tab.c ../core/coff.c ../core/exception.c ../core/macii_symbols.c ../core/redblack.c ../core/scsi.c ../core/video.c ../core/filesystem.c ../core/alloc_pool.c ../core/toby_frame_buffer.c ../core/ethernet.c ../core/sound.c sdl.c -lpthread -lm -lSDL2 -lGL -o shoebill
fpu.post.c: In function ‘_native_tentox’:
fpu.post.c:821:9: warning: type of ‘a’ defaults to ‘int’ [-Wimplicit-int]
821 | double _native_tentox(a) {
| ^~~~~~~~~~~~~~
/usr/bin/ld: /tmp/shoebill.lkIeQb.ltrans0.ltrans.o: in function main': /home/viktormadarasz/shoebill/sdl-gui/sdl.c:494: warning: sys_errlist' is deprecated; use strerror' or strerror_r' instead

[retokromer]

I presume, this repository is no longer maintained?

[tsal]

Check here: https://github.com/emaculation/shoebill

[retokromer]

Looking at the last commits, sadly it seems also to be only little maintained… I made a few changes, but I don’t know where to post a PR for the community.

[mitchblank]

This isn't Fedora-specific, it fails on modern clang compilers. Looking at core/fpu.c

#define _native_log(a) log(a)
#define _native_log1p(a) log((a) + 1.0) /* or log1p() */
double  _native_tentox(a) {

It's clear from context that a is meant to be a "double". I'd recommend this one-line change:

diff --git a/core/fpu.c b/core/fpu.c
index ce50a37..0277072 100644
--- a/core/fpu.c
+++ b/core/fpu.c
@@ -814,7 +814,7 @@ float128 _from_native(double n)
 #define _native_log2(a) (log(a) / log(2.0)) /* or log2() */
 #define _native_log(a) log(a)
 #define _native_log1p(a) log((a) + 1.0) /* or log1p() */
-double  _native_tentox(a) {
+static double _native_tentox(double a) {
     /*
      * This is a dumb workaround for a clang bug on OS X 10.10
      * Clang wants to optimize pow(10.0, a) to __exp(a), but