Experimental fork of SDCC with support for dead code elimination. https://gitea.privatedns.org/xavi/sdcc-gas/atom?h=master 2019-11-10T19:13:59+00:00 2019-11-10T19:13:59+00:00 Xavi Del Campo xavi.dcr@tutanota.com 2019-11-10T19:13:59+00:00 urn:sha1:40ce43b7aec0aaf96770412d041ede6bf0f0573d 2019-11-07T22:03:29+00:00 Xavi Del Campo xavi.dcr@tutanota.com 2019-11-07T22:03:29+00:00 urn:sha1:a13561bc9a51f646cabb32edf965af66dc769d56 2019-11-06T03:34:31+00:00 Xavi Del Campo xavi.dcr@tutanota.com 2019-11-06T03:21:13+00:00 urn:sha1:53953142e739085cc82a75dfd6b75b02e1e90eb0 2019-11-06T03:34:26+00:00 Xavi Del Campo xavi.dcr@tutanota.com 2019-11-06T03:10:53+00:00 urn:sha1:cf1d3aec552ebaea437935e74e5eabda4bd989e7 Issue #1 was caused because code_name and data_name had been incorrectly defined as const-qualified pointers, but this must be an exception for the z80 backend, as stated by the comments. Issue #2 has been solved by removing the --gas switch from SDCCmain.c, renaming it to "--asm=" (as requested by PKK) and moved to STM8's main.c, so the option is limited to those targets supporting it. Issue #4 was because of SDCCglue.c, that generated local labels using "!labeldef" instead of "!slabeldef", causing those __str_%d labels to be printed as "__str_%d::" (the double colons stand for globa label according to asxxx documentation) instead of "__str_%d":". I ran the regression tests and most of them passed, but "wchar" returned another "Multiple definition of..." error that I cannot explain so far. It is better than nothing, though. 2019-11-05T16:57:22+00:00 Xavi Del Campo xavi.dcr@tutanota.com 2019-11-05T16:57:22+00:00 urn:sha1:0290d8bd12f3cd188c200367ade7d569a92f1d2d 2019-11-05T16:56:35+00:00 Xavi Del Campo xavi.dcr@tutanota.com 2019-11-05T16:56:35+00:00 urn:sha1:e524f157266c07a7a4cf3b766edd6ff9db98e4f4 2019-10-29T23:23:41+00:00 Xavier ASUS xavi92psx@gmail.com 2019-10-29T23:23:41+00:00 urn:sha1:1dcefbf527e52315abfe3b4871272b3fd9bb0f4b 2019-10-29T23:04:10+00:00 Xavier ASUS xavi92psx@gmail.com 2019-10-29T23:04:10+00:00 urn:sha1:7c4ae5853b472f0a2bb1dd049a818ab8346c5cff 2019-10-28T01:03:38+00:00 Xavier ASUS xavi92psx@gmail.com 2019-10-28T01:03:38+00:00 urn:sha1:3378ab8f236f387978fa4cc75bd5fca7021e8a13 OTOH, section names included in "options" were being assigned *before* parsing command line arguments. These arose some issues so I have created a post-parsing function. 2019-10-25T00:56:56+00:00 Xavier ASUS xavi92psx@gmail.com 2019-10-25T00:56:56+00:00 urn:sha1:e801690e4fc177f063e2ae854b3f2a69bd1b9ba0 Memory addresses should be preceded by '#' when referring to them as integers