#3604 gte/20040307-1.c fails for hc08/s08 on big-endian host

[Philipp Klaus Krause]

According to the snapshots page, the regression test gte/20040307-1.c fails for macOS on PowerPC, which is currently the only big-endian system in the compile farm.

[Erik Petrich]

I can confirm that incorrect code is being generated:

_main_:
;/Users/sdcc-builder/build/sdcc-build/orig/sdcc/support/regression/cases/../gte/20040307-1.c:14: } sdata = {0x01};
;     genPointerSet
;     genPackBitsImmed
    lda _main__sdata_65536_8
    ora #0x00
    sta _main__sdata_65536_8

The "ora #0x00" should be "ora #0x01". However, if the underlying literal value was truly 0, the ora instruction should be omitted, which implies that the literal value is not actually 0 but somehow still converts to "00" with the %02x formatting specifier.

[Philipp Klaus Krause]

That looks like it could be the bug: %02x is a format specifier for an unsigned int. The argument is a signed int.

P.S.: Actually, litval is a unsigned long cast to signed int cast to unsigned long long. So %02llx should be correct. I'll put that into the next branch.

[Philipp Klaus Krause]

Fixed in the next branch in [r14194].

[Philipp Klaus Krause]

Fixed in [r14211] by merging next to trunk.