Tests: Assert failure in library-test.scm on MinGW

[Leonardo Valeri Manera]

"gcc" "a.c" -o "a.o" -c  -fno-strict-aliasing -fwrapv -DHAVE_CHICKEN_CONFIG_H -DC_ENABLE_PTABLES -march=native -O2 -IC:/Dev/MinGW32/include -IC:/Dev/GnuWin32/include -I.. -I"C:/Dev/Chicken/include/chicken"
rm a.c
"gcc" "a.o" -o "a.out" -Wl,--enable-auto-import -LC:/Dev/MinGW32/lib -LC:/Dev/GnuWin32/lib -L.. -L"C:/Dev/Chicken/lib" -lchicken -lm -lws2_32
rm a.o

Error: (library-tests.scm:204) assertion failed: (= (atan 42. 1.2) (fpatan2 42. 1.2))

Some naive testing of the above in csi and csc yields identical numbers from both functions. MinGW is lastest stable.

[Leonardo Valeri Manera]

Log of entire make check run.

[Leonardo Valeri Manera]

Since I was unclear in description, this is core @ ac0a3d132f772da5c543288af6d0d33f66c8fecb from Sat, June 30.

[Leonardo Valeri Manera]

As requested, .c output attached

[sjamaan]

This C output, when compiled with gcc on a BSD box and linked against master's libchicken runs and exits with exit code 0, without printing anything. This is the expected behavior, so the code generation is fine.

[Leonardo Valeri Manera]

I added

printf("%lf, %lf\n",  C_flonum_magnitude(t3), C_flonum_magnitude(t4));

to the else branch as per your request, and manually compiled using the gcc calls from the log:

c:\Dev\_src\chicken\trunk\tests>a.exe
-0.000000, 0.000000

Something seems off here ...

[sjamaan]

Strangely, the value printed by (print (atan 42. 1.2)) is 1.5422326689561365 in both interpreted and compiled mode...

[Leonardo Valeri Manera]

Problem caused by -O2, close?

[sjamaan]

Replying to Taverius:

Problem caused by -O2, close?

Unsure. Felix?

[sjamaan]

Never mind, we just discussed this and it turns out the -O2 is due to local tweaks to the MingW system that cause it to build with -O2; Chicken doesn't do that by default.

[felix winkelmann]

Milestone 4.8.0 deleted