scc, branch HEAD

cc1: Accept integer comparations in enum variables

2026-03-05T14:31:26Z

commit 9b162b0052316caaec77869017c395d0d050033e parent 05c02ab56769cd7ff85630d501627941d6abe84f Author: Roberto E. Vargas Caballero Date: Thu, 5 Mar 2026 15:31:26 +0100 cc1: Accept integer comparations in enum variables

tests: Update README

2026-02-20T14:04:53Z

commit 05c02ab56769cd7ff85630d501627941d6abe84f parent 494b5e954619ec4212db5a08c3f3b355b8104785 Author: Roberto E. Vargas Caballero Date: Fri, 20 Feb 2026 15:04:53 +0100 tests: Update README

cc1: Fix va_start() warning

2026-02-20T08:32:02Z

commit 494b5e954619ec4212db5a08c3f3b355b8104785 parent 9ef71d0377a399f0e0c8e3bcfd47843fb5d66210 Author: Roberto E. Vargas Caballero Date: Fri, 20 Feb 2026 09:32:02 +0100 cc1: Fix va_start() warning The va_start() builtin was checking if the parameter actually received was the actual last named parameter of a function with variable argument list, but it tried to get the ellipsis type from the list of arguments itself, but the list of arguments only has the named arguments, so it was not possible to match the ellipsis type. In order to check that it is required to go to the list of argument types of the function type.

cc1: Mark builltin with side effect

2026-02-20T08:23:49Z

commit 9ef71d0377a399f0e0c8e3bcfd47843fb5d66210 parent 99a2aaeb775c3ba5d69e48b2501af71fce63e648 Author: Roberto E. Vargas Caballero Date: Fri, 20 Feb 2026 09:23:49 +0100 cc1: Mark builltin with side effect The vaarg builtins are used to handle variable length argument and they do modify the va argument that they receive, so emiting warnings for them didn't make sense.

build: Use LIBPREFIX for crt files in musl

2026-02-18T20:01:47Z

commit 99a2aaeb775c3ba5d69e48b2501af71fce63e648 parent 28c274e634e5bb1a8a99ee403f1373b6db7fcd07 Author: Roberto E. Vargas Caballero Date: Wed, 18 Feb 2026 21:01:47 +0100 build: Use LIBPREFIX for crt files in musl In the case of musl there is a total disconnection between the prefix used by scc and the prefix used when building and installing musl. In order to make easier the configuration of musl this commits just insert the prefix obtained by the output of gcc.

cc1: Warn only in different macro redefinitions

2026-02-18T18:45:06Z

commit 28c274e634e5bb1a8a99ee403f1373b6db7fcd07 parent 40275b12f4b192d63d7d6702fd3caaa8eb3f5bc5 Author: Roberto E. Vargas Caballero Date: Wed, 18 Feb 2026 19:45:06 +0100 cc1: Warn only in different macro redefinitions The standard says that redefining a macro with a different value is undefined behaviour. Scc uses to warn in any redefinition of macros, but as there are some projects that like to redefine them continously this generated too much noise, and this commit moved the warning to only when the defintion is different.

doc: Update scc man page to scc-cc

2026-02-18T09:18:32Z

commit 40275b12f4b192d63d7d6702fd3caaa8eb3f5bc5 parent 7c107f7208d33678855fb5ff8197d2a085cc4617 Author: Roberto E. Vargas Caballero Date: Wed, 18 Feb 2026 10:18:32 +0100 doc: Update scc man page to scc-cc Scc is the wrapper tool now, and the name of the compiler binary is scc-cc.

driver/posix: Improve command line compatibility

2026-02-18T09:16:48Z

commit 7c107f7208d33678855fb5ff8197d2a085cc4617 parent 42b50c811c8ad0f6db4fca6088bc67f33cce224e Author: Roberto E. Vargas Caballero Date: Wed, 18 Feb 2026 10:16:48 +0100 driver/posix: Improve command line compatibility Sadly, the world is out of any hope, and accepting to ignore flags used by some mainstream compilers help a lot.

build: Use better config-musl target

2026-02-16T15:53:25Z

commit 42b50c811c8ad0f6db4fca6088bc67f33cce224e parent 5042ac1a04b0b0e797f20fb56d9e377870d513c7 Author: Roberto E. Vargas Caballero Date: Mon, 16 Feb 2026 16:53:25 +0100 build: Use better config-musl target The config-musl target was a helper to make easier to configure scc with a musl external libc, but it failed in native musl distributions like for example alpine. This new target is expected to be more robust and work in native and non native distributions.

cc1: Add unary - folding for floats

2026-02-16T15:46:00Z

commit 5042ac1a04b0b0e797f20fb56d9e377870d513c7 parent 8b3404606cd8120322ef7816442a7fb938e7b513 Author: Roberto E. Vargas Caballero Date: Mon, 16 Feb 2026 16:46:00 +0100 cc1: Add unary - folding for floats This is required for static initializers.

cc1: Mark cpp else blocks

2026-02-16T10:24:54Z

commit 8b3404606cd8120322ef7816442a7fb938e7b513 parent dd2b510053f9b9bfd2eb4f249d493b5f42c3c489 Author: Roberto E. Vargas Caballero Date: Mon, 16 Feb 2026 11:24:54 +0100 cc1: Mark cpp else blocks Once we find a #else then no more #else or #elif can happen in the current preprocessor conditional context and we have to mark it to be able to detect this condition.

tests/cc: Use diff -u

2026-02-16T10:07:29Z

commit dd2b510053f9b9bfd2eb4f249d493b5f42c3c489 parent 3300cf3d3e2a83b6b6c438fd40b8a9a6b164900e Author: Roberto E. Vargas Caballero Date: Mon, 16 Feb 2026 11:07:29 +0100 tests/cc: Use diff -u Diff -u generates diff files easier to understand.

libc: Improve gcc-ssc

2026-02-16T09:21:16Z

commit 3300cf3d3e2a83b6b6c438fd40b8a9a6b164900e parent d1f3dc2a94c8cb46bb67259642a1059983de36be Author: Roberto E. Vargas Caballero Date: Mon, 16 Feb 2026 10:21:16 +0100 libc: Improve gcc-ssc The crt file should be the first object in the list, because it pulls the main symbol and maybe other required symbols. For many platforms it is not a big issue because the crt only uses the main symbol, but in some other cases it may pull other symbols. Also, forcing -g didn't make too much sense, and it was not difficult to handle it correctly.

doc: Update LICENSE file

2026-02-16T09:09:47Z

commit d1f3dc2a94c8cb46bb67259642a1059983de36be parent 8b8b25fa24223ca929a77253d26c76b3c596ffd6 Author: Roberto E. Vargas Caballero Date: Mon, 16 Feb 2026 10:09:47 +0100 doc: Update LICENSE file

gcc-scc.sh: parse options using getopts

2026-02-15T20:42:44Z

commit 8b8b25fa24223ca929a77253d26c76b3c596ffd6 parent c31a24c04de5785a76725ea7e2fea652e192b107 Author: Avi Halachmi (:avih) Date: Sun, 15 Feb 2026 22:42:44 +0200 gcc-scc.sh: parse options using getopts Previously, the options parsing code rejected some valid inputs, such as -ofile (in one arg). Now we use standarg getopts which conforms to the POSIX syntax guidelines. Additionally, improve the error message on unknown option: - print to stderr instead of stdout. - quote it (otherwise these are globs...). - Add spaces between the options (still less than 70 chars).

scc-cpp.sh and gcc-scc.sh: improve handling of arguments

2026-02-15T20:22:19Z

commit c31a24c04de5785a76725ea7e2fea652e192b107 parent 534a9fa01a8a0c080df60800a370421dc594532c Author: Avi Halachmi (:avih) Date: Sun, 15 Feb 2026 22:22:19 +0200 scc-cpp.sh and gcc-scc.sh: improve handling of arguments Previously, any argument to scc-cpp.sh and some arguments to gcc-scc.sh were broken if they contained any IFS chars, such as space. Now correct quoting is used in these scripts, which do maintain the arguments correctly. In scc-cpp.sh the change is trivial from $@ to "$@" . In gcc-scc.sh it's a bit more involved as arguments are handled individually and iteratively. The big change is replacing .c with .o for all file names for $ld, where instead of converting the whole line in one sed, we now iterate the files individually while maintaining any IFS chars in them

make: Escape values passed in MAKEFLAGS

2026-02-15T10:12:28Z

commit 534a9fa01a8a0c080df60800a370421dc594532c parent d94c6547755e253374bb6da489b5c01f69680dbc Author: Roberto E. Vargas Caballero Date: Sun, 15 Feb 2026 11:12:28 +0100 make: Escape values passed in MAKEFLAGS The POSIX standard says: The macro= value macro definition operands can also be included. The difference between the contents of MAKEFLAGS and the make utility command line is that the contents of the variable shall not be subjected to the word expansions (see Word Expansions) associated with parsing the command line values. And it basically means that we have to escape the parameters passed in MAKEFLAGS and then unescape and doing the word spliting. The code had a naive strtok() implementation for word spliting, and it had to be replaced for a more complex parsing function which also handles escapes and quotes with ' and ".

posix: Quote parameters passed down in scc

2026-02-15T10:07:41Z

commit d94c6547755e253374bb6da489b5c01f69680dbc parent ffc56872a365f518ba8899157a2738a5462df500 Author: Roberto E. Vargas Caballero Date: Sun, 15 Feb 2026 11:07:41 +0100 posix: Quote parameters passed down in scc The scc wrapper has to preserve the parameters in the best possible way to avoid spliting words in the wrong places when they are passed down to the specific tools.

cc2: Add missed dependencies for riscv

2026-02-12T11:46:45Z

commit ffc56872a365f518ba8899157a2738a5462df500 parent 0dc121edaac1b54277f2eac7c3fac45746326ff2 Author: Roberto E. Vargas Caballero Date: Thu, 12 Feb 2026 12:46:45 +0100 cc2: Add missed dependencies for riscv

tests/make: Fix mistakes after trap changes

2026-02-11T13:44:44Z

commit 0dc121edaac1b54277f2eac7c3fac45746326ff2 parent 5eedaa9f030481f63ab828c89c0623068fd2b7a7 Author: Roberto E. Vargas Caballero Date: Wed, 11 Feb 2026 14:44:44 +0100 tests/make: Fix mistakes after trap changes

make: Change default CFLAGS and FFLAGS

2026-02-11T13:38:41Z

commit 5eedaa9f030481f63ab828c89c0623068fd2b7a7 parent 3ef4da68f555260111abb67615a28885bf57a7b5 Author: Roberto E. Vargas Caballero Date: Wed, 11 Feb 2026 14:38:41 +0100 make: Change default CFLAGS and FFLAGS This is a very weird situation, because classical cc accepted -O to enable optimizations, but POSIX 2004 specified -O level for c99, but it specified -O for default value of CFLAGS in make. This conjuction made that gcc accepts -O1 but not -O 1, because -O 1 is interpreted like the -O option and the source file 1. That makes that the default CFLAGS of POSIX make cannot work with gcc (and clang), but GNU make uses -O1 instead of the -O 1 specified in POSIX. Seems both values should be accepted by compliant POSIX tools, in order to avoid problems the best option is to follow GNU make here and use -O1 instead of -O 1, even when it is not strictly POSIX compliant.

cc2: Generalize keeplabel()

2026-02-10T13:38:36Z

commit 3ef4da68f555260111abb67615a28885bf57a7b5 parent 61adffcdcf6e9aeb0c762fb0ba3974d6c2e648c4 Author: Roberto E. Vargas Caballero Date: Tue, 10 Feb 2026 14:38:36 +0100 cc2: Generalize keeplabel() Having a keeplabel() call in every case where we can modify the statement with the current label is very fragile, and we should to search for a more general solution where we don't have to care anymore about this problem. We could modify the parser to get rid of labels in statements and have only label statements, but it is much more simple to check for that case in every call to sethi().

cc2: Keep label position in branches

2026-02-10T10:26:43Z

commit 61adffcdcf6e9aeb0c762fb0ba3974d6c2e648c4 parent 91d6c4c2e9cf507de205fab129a65be40fb1c6b3 Author: Roberto E. Vargas Caballero Date: Tue, 10 Feb 2026 11:26:43 +0100 cc2: Keep label position in branches When a branch is transformed into a set of jumps new statements are inserted before the current statement, displacing the relative position of the label of the current statement in case of having one. For that reason in this case we have to create a label statment and place there the label.

cc2: Keep label position in commas

2026-02-10T10:26:43Z

commit 91d6c4c2e9cf507de205fab129a65be40fb1c6b3 parent 9e85b6a1ad15b70aefd3436ca605aea2f523f335 Author: Roberto E. Vargas Caballero Date: Tue, 10 Feb 2026 11:26:43 +0100 cc2: Keep label position in commas When a comma is transformed into a set of jumps new statements are inserted before the current statement, displacing the relative position of the label of the current statement in case of having one. For that reason in this case we have to create a label statment and place there the label.

cc2: Keep label position in logic expressions

2026-02-10T10:26:43Z

commit 9e85b6a1ad15b70aefd3436ca605aea2f523f335 parent 8be6d9e4cd588fd6110a4252e36848214ad04b28 Author: Roberto E. Vargas Caballero Date: Tue, 10 Feb 2026 11:26:43 +0100 cc2: Keep label position in logic expressions When a logic is transformed into a set of jumps new statements are inserted before the current statement, displacing the relative position of the label of the current statement in case of having one. For that reason in this case we have to create a label statment and place there the label.

cc2: Keep label position in ternaries

2026-02-10T10:26:43Z

commit 8be6d9e4cd588fd6110a4252e36848214ad04b28 parent f3c4d6c3e0822ccf373935eb8851660bbd298931 Author: Roberto E. Vargas Caballero Date: Tue, 10 Feb 2026 11:26:43 +0100 cc2: Keep label position in ternaries When a ternary is transformed into a set of jumps new statements are inserted before the current statement, displacing the relative position of the label of the current statement in case of having one. For that reason in this case we have to create a label statment and place there the label.

cc2: Preserve labels in replace()

2026-02-10T10:00:03Z

commit f3c4d6c3e0822ccf373935eb8851660bbd298931 parent 46588de4ca1b0a4c47591d61c5947dc1a3238ce5 Author: Roberto E. Vargas Caballero Date: Tue, 10 Feb 2026 11:00:03 +0100 cc2: Preserve labels in replace() The intention of replace is replacing a node with other, but in case the original node had a label it was lost in the replace process.

cc1: Add support for non constant auto initializers

2026-02-09T15:08:49Z

commit 46588de4ca1b0a4c47591d61c5947dc1a3238ce5 parent 34b628de0fe4f297e8e5ef969a6355748a8154ef Author: Roberto E. Vargas Caballero Date: Mon, 9 Feb 2026 16:08:49 +0100 cc1: Add support for non constant auto initializers This has been an important miss in scc for a while. While this code is not guranteed to work for all the cases, and it can generate really bad code, it works for the most common casesa and it can be used to bootstrap them.

cc1: Disable string emit in initializers

2026-02-04T11:41:41Z

commit 34b628de0fe4f297e8e5ef969a6355748a8154ef parent c74e72c72af0f7ce01fe1388f4d542a57c5ed70f Author: Roberto E. Vargas Caballero Date: Wed, 4 Feb 2026 12:41:41 +0100 cc1: Disable string emit in initializers When we are dealing with initializers there are transformations done to the original strings to fit with the target type, like adding additional 0 or removing characters if it is bigger. That meant that we were emitting strings that were not actually used, because the initializers were emitting their own copy of these strings. In case of having an error in initialize() the disstring variable would remain set, but that is not a problem because in that state we don't emit anything else. Also, the function str2ary() had a bug hidden in how the new string was generated, but due to other bug that string was not used ever. The new code avoids using str* alike functions because it would break when the array has embedded NUL characters.

tests/cc: Add 0239-fcasts.c

2026-02-02T11:43:35Z

commit c74e72c72af0f7ce01fe1388f4d542a57c5ed70f parent cdd599b1db41f2983c446c0fdefb799afbb511eb Author: Roberto E. Vargas Caballero Date: Mon, 2 Feb 2026 12:43:35 +0100 tests/cc: Add 0239-fcasts.c

cc1: Fix padding in initializers

2026-02-02T11:37:24Z

commit cdd599b1db41f2983c446c0fdefb799afbb511eb parent 42441dfa818cce4c67092948f8113fd1afef41ab Author: Roberto E. Vargas Caballero Date: Mon, 2 Feb 2026 12:37:24 +0100 cc1: Fix padding in initializers The alignment calculation was wrong and it generated static struct and arrays with data misplaced.

objdump: Minor typo in error string

2026-01-31T19:36:23Z

commit 42441dfa818cce4c67092948f8113fd1afef41ab parent 8287e1aeccdcf88a7ac6d415ca162ccb474da3af Author: Roberto E. Vargas Caballero Date: Sat, 31 Jan 2026 20:36:23 +0100 objdump: Minor typo in error string

cc2/qbe: Fix floating point conversions

2026-01-31T18:53:02Z

commit 8287e1aeccdcf88a7ac6d415ca162ccb474da3af parent 0f0d3edeb1988957d0196482eec84366956cb574 Author: Roberto E. Vargas Caballero Date: Sat, 31 Jan 2026 19:53:02 +0100 cc2/qbe: Fix floating point conversions The operations were inverted and it tried to truncate when it had to extend, and it tried to extend when it had to truncate.

cc1: Fold constant casts involving floats

2026-01-31T18:05:39Z

commit 0f0d3edeb1988957d0196482eec84366956cb574 parent 3b5e9eb7079be6f946a66c2b751921723680a5aa Author: Roberto E. Vargas Caballero Date: Sat, 31 Jan 2026 19:05:39 +0100 cc1: Fold constant casts involving floats

cc2/qbe: Fix bug introduced in b83ec8ed

2026-01-31T08:50:03Z

commit 3b5e9eb7079be6f946a66c2b751921723680a5aa parent eaf2b18533e87462e71e588fd6edd96d937622b2 Author: Roberto E. Vargas Caballero Date: Sat, 31 Jan 2026 09:50:03 +0100 cc2/qbe: Fix bug introduced in b83ec8ed The commit b83ec8ed modified the code of binary operations because in the case of comparisions we have a cmp node with float type, but the result is integer. The intermediate language used cannot express cases like this one, because it assumes that the operand and the result have the same type. The commit b83ec8ed assumes that the piece of code modified applied only to comparisions, but in fact, it affected to any binary operation, producing that all the binary operations were casted to int32type.

cc1: Fix definition of __FLT_EVAL_METHOD__

2026-01-30T20:50:42Z

commit eaf2b18533e87462e71e588fd6edd96d937622b2 parent b83ec8ed7ab5e20a968b8e1a00236036ea363131 Author: Roberto E. Vargas Caballero Date: Fri, 30 Jan 2026 21:50:42 +0100 cc1: Fix definition of __FLT_EVAL_METHOD__

cc1: Use type punning to pass floats to cc2

2026-01-30T20:37:48Z

commit b83ec8ed7ab5e20a968b8e1a00236036ea363131 parent f41102032108467135842e9f36c4dfd1072c4d62 Author: Roberto E. Vargas Caballero Date: Fri, 30 Jan 2026 21:37:48 +0100 cc1: Use type punning to pass floats to cc2 Doesn't mind what we do that we cannot pass a floating point number from cc1 to cc2 in a fully portable way, and using type punning at least maintain the endianess correctly. This code should even work in cross compilation as far as the host and the target use the same floating point representation, that today will likely be the ieee 754.

cc1: Fix pointer cast type

2026-01-30T16:06:37Z

commit f41102032108467135842e9f36c4dfd1072c4d62 parent 475fa03250d04142fb1aeb9635ffe7e3eac434c8 Author: Roberto E. Vargas Caballero Date: Fri, 30 Jan 2026 17:06:37 +0100 cc1: Fix pointer cast type

cc1: Add support for float constant

2026-01-30T14:44:17Z

commit 475fa03250d04142fb1aeb9635ffe7e3eac434c8 parent c3b8a3ef1ef7df8ae85b1d0b9061725a02c356dc Author: Roberto E. Vargas Caballero Date: Fri, 30 Jan 2026 15:44:17 +0100 cc1: Add support for float constant Strtold already manages all the complexity of dealing with floating point numbers.

cc1: Add support for c99 float arithmetic

2026-01-30T11:49:24Z

commit c3b8a3ef1ef7df8ae85b1d0b9061725a02c356dc parent b057d1fdf510c655916a99184421b0197c5da056 Author: Roberto E. Vargas Caballero Date: Fri, 30 Jan 2026 12:49:24 +0100 cc1: Add support for c99 float arithmetic C99 allows arithmetic operations to be performed in floats, while K&R and C90 promoted floats to double. This was the behaviour of scc until now, but this commit enables operations in float and also add folding of long double constants.

cc1: Be careful about float optimizations

2026-01-30T09:42:53Z

commit b057d1fdf510c655916a99184421b0197c5da056 parent b8ad553523ba96ddf4c363befb1a5397b18af440 Author: Roberto E. Vargas Caballero Date: Fri, 30 Jan 2026 10:42:53 +0100 cc1: Be careful about float optimizations As described in 5.1.2.3p14: Rearrangement for floating-point expressions is often restricted because of limitations in precision as well as range. The implementation cannot generally apply the mathematical associative rules for addition or multiplication, nor the distributive rule, because of roundoff error, even in the absence of overflow and underflow. Likewise, implementations cannot generally replace decimal constants in order to rearrange expressions. In the following fragment, rearrangements suggested by mathematical rules for real numbers are often not valid.

cc1: Fold correctly unary in sizeof

2026-01-29T19:53:34Z

commit b8ad553523ba96ddf4c363befb1a5397b18af440 parent 7dbb2d53125f9592cf579c0e948c56eb886a8a3c Author: Roberto E. Vargas Caballero Date: Thu, 29 Jan 2026 20:53:34 +0100 cc1: Fold correctly unary in sizeof When a parenthesis happens after a sizeof it can mean a type name or an unary expression. The code was correctly detecting the type name case, but in the unary case it assumed that the unary expresion was equal to ( expr ) but unary can be also ( expr ) postfix, failling in cases like: sizeof(array)[0]

cc1: Use int for namespaces

2026-01-29T19:22:24Z

commit 7dbb2d53125f9592cf579c0e948c56eb886a8a3c parent 8e07fdb59f3eaca0e5e8ad160f4d2ae37270360e Author: Roberto E. Vargas Caballero Date: Thu, 29 Jan 2026 20:22:24 +0100 cc1: Use int for namespaces The code in decl.c was protecting against overflow using INT_MAX, but the Symbols and types were using unsigned char or char, that made mismatching when the lookup was performed.

cc1: Parse correctly **/ comments

2026-01-29T14:48:18Z

commit 8e07fdb59f3eaca0e5e8ad160f4d2ae37270360e parent 661b5b01217b553c0e0b142ae41bdf9c9b2ac14b Author: Roberto E. Vargas Caballero Date: Thu, 29 Jan 2026 15:48:18 +0100 cc1: Parse correctly **/ comments The parser was consuming always a character when a * was found, making that even number of * followed by a / were not considered an end of comment.

cc1: Accept digits in macro parameters

2026-01-29T12:35:25Z

commit 661b5b01217b553c0e0b142ae41bdf9c9b2ac14b parent b04e1786f2fbbf3976b6d0aba4cc122aefd23bc7 Author: Roberto E. Vargas Caballero Date: Thu, 29 Jan 2026 13:35:25 +0100 cc1: Accept digits in macro parameters The last changes about parsing macro definitions creates a new bug because they defined identifiers only using alpha and _ characters, but it should accept digits after the 1st character.

cc1: Fix elif handling

2026-01-29T11:36:50Z

commit b04e1786f2fbbf3976b6d0aba4cc122aefd23bc7 parent 740fd29faa97cc8a39c7e48c368191bc2df9cc99 Author: Roberto E. Vargas Caballero Date: Thu, 29 Jan 2026 12:36:50 +0100 cc1: Fix elif handling The code was assuming that an #else was always inverting the state of the preprocessor output, and elif was handled like a #else plus a #if, but that was not the actual case. The code tried to manage this situation counting the number of nested enabling/disabling of cppoff, but it was a non sense because we already had a stack of cpp contexts that could do the work much better. Also, we needed a sticky variable (called "done" in this commit) to track that one case of the if-else chain was true before, because the "enabled" field changes and then we forget the history of the chain.

cc1: Simplify getdefs()

2026-01-29T11:25:29Z

commit 740fd29faa97cc8a39c7e48c368191bc2df9cc99 parent 58a4f1584eb24dd4204d8b4d87b06a11acede509 Author: Roberto E. Vargas Caballero Date: Thu, 29 Jan 2026 12:25:29 +0100 cc1: Simplify getdefs() We can unify the IDEN and string cases if we use yytext/yylen in the case of IDEN because the STRING case is handled by the tokenizer to avoid code duplication.

cc1: Remove macro tokens from the tokenizer

2026-01-28T14:12:29Z

commit 58a4f1584eb24dd4204d8b4d87b06a11acede509 parent dcde10bc7be69fa1c7e497f67754f5323649f5cf Author: Roberto E. Vargas Caballero Date: Wed, 28 Jan 2026 15:12:29 +0100 cc1: Remove macro tokens from the tokenizer The macro definitions are currently handled by the preprocessor itself which means that the tokenizer should not see these tokens anymore in a valid sequence, and it means that we can remove it.

cc1: Rewrite macro defintion parser

2026-01-28T13:58:35Z

commit dcde10bc7be69fa1c7e497f67754f5323649f5cf parent fad9e0fb2e77d701992ea2e740f8c3bb67e9a049 Author: Roberto E. Vargas Caballero Date: Wed, 28 Jan 2026 14:58:35 +0100 cc1: Rewrite macro defintion parser The macro definition parser was based in the tokenizer used by cc1 but it means that the input string was modified adding spaces around the tokens, which in normal use cases does not generate any problems. Some people does things like: #define STDIO #include STDIO that is not conformant becuase the preprocessor is allowed to work based in C tokens that would split the previous macro in: < stdio . h > This commits modifies the parser of macro definitions to not using the tokenizer (except in the case of strings) and preserve the input string as much as possible.

libc: Add missed bits/wchar.h to the proto file

2026-01-27T12:12:06Z

commit fad9e0fb2e77d701992ea2e740f8c3bb67e9a049 parent fbf964d8de9d5dc603d674e67f9c1236373827e6 Author: Roberto E. Vargas Caballero Date: Tue, 27 Jan 2026 13:12:06 +0100 libc: Add missed bits/wchar.h to the proto file

libc: Protect redefinitions of __va_list and __wchar

2026-01-27T11:11:06Z

commit fbf964d8de9d5dc603d674e67f9c1236373827e6 parent b392aeafa794ca5b5aecdf93475c3869af3a9747 Author: Roberto E. Vargas Caballero Date: Tue, 27 Jan 2026 12:11:06 +0100 libc: Protect redefinitions of __va_list and __wchar

cc1: Remove NR_MAXSTRUCTS

2026-01-27T08:55:06Z

commit b392aeafa794ca5b5aecdf93475c3869af3a9747 parent 0b09e68a1bc71be5eaa5754c98365cf677992aed Author: Roberto E. Vargas Caballero Date: Tue, 27 Jan 2026 09:55:06 +0100 cc1: Remove NR_MAXSTRUCTS This limit is not defined by any standard, and it was creating an artificial limit that does not help at all to discover non portable programs because it checked for something useless.

Merge remote-tracking branch 'origin'

2026-01-24T10:51:44Z

commit 0b09e68a1bc71be5eaa5754c98365cf677992aed parent dc99102efafb82f90ae3f39ce98f30e06f94335b Author: Roberto E. Vargas Caballero Date: Sat, 24 Jan 2026 11:51:44 +0100 Merge remote-tracking branch 'origin'

tests: Avoid using mktemp

2026-01-24T10:50:19Z

commit dc99102efafb82f90ae3f39ce98f30e06f94335b parent 68568d529414af1522832b6fe63bf163306e0c8f Author: Roberto E. Vargas Caballero Date: Sat, 24 Jan 2026 11:50:19 +0100 tests: Avoid using mktemp Mktemp is not posix and it is usually not needed for our use case.

build: Use exit $? in signal traps

2026-01-24T10:47:43Z

commit 68568d529414af1522832b6fe63bf163306e0c8f parent a4f6e0ebf74b40d95b8d4de20bd0eb890efefaf3 Author: Roberto E. Vargas Caballero Date: Sat, 24 Jan 2026 11:47:43 +0100 build: Use exit $? in signal traps This helps to preserve the exit status instead of just using 1 for all the signals.

cc1: Fix comment typos

2026-01-16T12:15:44Z

commit 8e32d090646d3eddcd4206c11920afc3a2c89137 parent cd97844f7cd875616f4b97c675b08ce41aee9df3 Author: Zhaoming Luo Date: Fri, 16 Jan 2026 12:15:44 +0000 cc1: Fix comment typos

doc: Fix typo in README

2026-01-16T12:15:42Z

commit cd97844f7cd875616f4b97c675b08ce41aee9df3 parent 52cf2e9986d52e8e812a30b94dc118dabb1aea1f Author: Zhaoming Luo Date: Fri, 16 Jan 2026 12:15:42 +0000 doc: Fix typo in README

build: Preserve exit status scripts/config

2026-01-06T00:14:33Z

commit a4f6e0ebf74b40d95b8d4de20bd0eb890efefaf3 parent 52cf2e9986d52e8e812a30b94dc118dabb1aea1f Author: Roberto E. Vargas Caballero Date: Tue, 6 Jan 2026 01:14:33 +0100 build: Preserve exit status scripts/config

libmach: Add findseg()

2026-01-05T21:22:18Z

commit 52cf2e9986d52e8e812a30b94dc118dabb1aea1f parent ece7fade7d459523af266428e95b4f587a2f0ba9 Author: Roberto E. Vargas Caballero Date: Mon, 5 Jan 2026 22:22:18 +0100 libmach: Add findseg() This function returns the segment index in the map. This function is mainly designed to be used in elf where there is a difference between sections and segments.

tests/libc: Add 0067-wcwidth

2026-01-05T20:33:03Z

commit ece7fade7d459523af266428e95b4f587a2f0ba9 parent 361a63ace0c9cec71586b8c6f79397a981c32589 Author: Roberto E. Vargas Caballero Date: Mon, 5 Jan 2026 21:33:03 +0100 tests/libc: Add 0067-wcwidth

objdump: Fix format string in dumpscns()

2026-01-05T20:19:16Z

commit 361a63ace0c9cec71586b8c6f79397a981c32589 parent 0f88acf52504636f56b73132285c7355b1beb0af Author: Roberto E. Vargas Caballero Date: Mon, 5 Jan 2026 21:19:16 +0100 objdump: Fix format string in dumpscns()

libc: Include a mbstate_t in FILE

2026-01-05T19:24:01Z

commit 0f88acf52504636f56b73132285c7355b1beb0af parent 84128eaf37e33fed023f271018b157b7398ad91c Author: Roberto E. Vargas Caballero Date: Mon, 5 Jan 2026 20:24:01 +0100 libc: Include a mbstate_t in FILE Several of the file operations related to wchar require a mbstate_t, and every stream should have its own mbstate_t and not using the hidden state of several of the conversion functions.

libc/stdlib: Minor improvements to realloc()

2026-01-04T20:18:49Z

commit 84128eaf37e33fed023f271018b157b7398ad91c parent ae0058502d81d2742783b1d1800784319341ff12 Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 21:18:49 +0100 libc/stdlib: Minor improvements to realloc()

libc/wchar: Whitespace change

2026-01-04T20:14:43Z

commit ae0058502d81d2742783b1d1800784319341ff12 parent 117e2caee5ca7b1441e2e4c4135c2aaac7b5cc78 Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 21:14:43 +0100 libc/wchar: Whitespace change

tests: Minor improvements

2026-01-04T20:09:03Z

commit 117e2caee5ca7b1441e2e4c4135c2aaac7b5cc78 parent 0f61a6528784b5add0eef1d64ea16eb5adf689f9 Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 21:09:03 +0100 tests: Minor improvements

libc/string: Minor improvements

2026-01-04T19:48:34Z

commit 0f61a6528784b5add0eef1d64ea16eb5adf689f9 parent 9a60d2a80efcd85c85f90b1c0406c8fe3f324f5b Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 20:48:34 +0100 libc/string: Minor improvements

libc/wchar: Add wctob()

2026-01-04T19:29:39Z

commit 9a60d2a80efcd85c85f90b1c0406c8fe3f324f5b parent c4b7399c26bf3040f33b4760e162495d74d85d73 Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 20:29:39 +0100 libc/wchar: Add wctob()

libc/wchar: Add btowc()

2026-01-04T19:17:51Z

commit c4b7399c26bf3040f33b4760e162495d74d85d73 parent 534e0b91e27ad3cb37404d696cd1c0044a8ebe18 Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 20:17:51 +0100 libc/wchar: Add btowc()

tests/libc: Add 0043-wcsrtombs

2026-01-04T18:47:25Z

commit 534e0b91e27ad3cb37404d696cd1c0044a8ebe18 parent 005f4c376c05bd22e21761cf063ea76fb9338dc1 Author: Roberto E. Vargas Caballero Date: Sun, 4 Jan 2026 19:47:25 +0100 tests/libc: Add 0043-wcsrtombs

libc: Add wcstok

2026-01-03T17:55:53Z

commit 005f4c376c05bd22e21761cf063ea76fb9338dc1 parent b2eb6e5269977992d88d388ff30a078aa9101a42 Author: Roberto E. Vargas Caballero Date: Sat, 3 Jan 2026 18:55:53 +0100 libc: Add wcstok

libc: Reorder common-objs.mk

2026-01-03T17:37:44Z

commit b2eb6e5269977992d88d388ff30a078aa9101a42 parent 836ed99e5aa87033470b3594597070cb0e00148b Author: Roberto E. Vargas Caballero Date: Sat, 3 Jan 2026 18:37:44 +0100 libc: Reorder common-objs.mk

Merge remote-tracking branch 'origin/master'

2026-01-03T16:47:00Z

commit 836ed99e5aa87033470b3594597070cb0e00148b parent 452190b652a22d2d5ba1fbc2106bf7847944f582 Author: Roberto E. Vargas Caballero Date: Sat, 3 Jan 2026 17:47:00 +0100 Merge remote-tracking branch 'origin/master'

build: Build bootstrap make in scripts

2026-01-03T14:35:47Z

commit 47361f9afa1e1b5b31f58052d3253e1067dcd773 parent 5b69547863e7e3da563e26164c34f22585920377 Author: Roberto E. Vargas Caballero Date: Sat, 3 Jan 2026 15:35:47 +0100 build: Build bootstrap make in scripts Just to keep the top level directory cleaner.

tests: Use $TEST in every scc invocation

2026-01-03T14:24:55Z

commit 5b69547863e7e3da563e26164c34f22585920377 parent 6d67d778d8a2bfe2093233d1fae81663f712aa2a Author: Roberto E. Vargas Caballero Date: Sat, 3 Jan 2026 15:24:55 +0100 tests: Use $TEST in every scc invocation using a variable before executing any of the commands allows to use emulators of checkers easilly just setting that variable.

build: Fix use of trap

2026-01-03T13:14:58Z

commit 6d67d778d8a2bfe2093233d1fae81663f712aa2a parent c825c03e6f3b1e3a757a47c75d77ce57ebb1624d Author: Roberto E. Vargas Caballero Date: Sat, 3 Jan 2026 14:14:58 +0100 build: Fix use of trap Using the same trap for EXIT and signals has the problem that signals don't finish the current execution. For that reason is better to use different traps and do the cleaning only in the EXIT trap while the signals just use exits that eventually will call the EXIT trap.

include: Remove parameter names in prototypes

2025-11-18T15:47:52Z

commit 452190b652a22d2d5ba1fbc2106bf7847944f582 parent 87e4396fb51faf4eac6fbd0a853e8c9d93c88fa2 Author: Roberto E. Vargas Caballero Date: Tue, 18 Nov 2025 16:47:52 +0100 include: Remove parameter names in prototypes Just be consistent and apply the same style in all the files.

include: Remove extern in function declarations

2025-11-18T14:26:53Z

commit 87e4396fb51faf4eac6fbd0a853e8c9d93c88fa2 parent ad5bd51963863fe5a3c3ad58a1ae42d578aa8980 Author: Roberto E. Vargas Caballero Date: Tue, 18 Nov 2025 15:26:53 +0100 include: Remove extern in function declarations This was introduced some time ago to follow some style but it does not help so much, and it only makes longer the declaration of the functions.

include: Remove T* types

2025-11-18T13:01:45Z

commit ad5bd51963863fe5a3c3ad58a1ae42d578aa8980 parent c825c03e6f3b1e3a757a47c75d77ce57ebb1624d Author: Roberto E. Vargas Caballero Date: Tue, 18 Nov 2025 14:01:45 +0100 include: Remove T* types These types didn't work very well, and at some point we stopped using them. Removing them makes the code more straigh forward and avoids some stupid casts in printf alike function calls.

doc: Remove outdated bug

2025-11-18T10:46:36Z

commit c825c03e6f3b1e3a757a47c75d77ce57ebb1624d parent 3d92a3b0d83de06ce99f320906711d2d21f9a0c4 Author: Roberto E. Vargas Caballero Date: Tue, 18 Nov 2025 11:46:36 +0100 doc: Remove outdated bug

tests/ar: Fix file name in 0039-update.sh

2025-10-31T17:19:56Z

commit 3d92a3b0d83de06ce99f320906711d2d21f9a0c4 parent 1ed0ff0000999561feee336c289252faf2502a7e Author: Roberto E. Vargas Caballero Date: Fri, 31 Oct 2025 18:19:56 +0100 tests/ar: Fix file name in 0039-update.sh

tests: Set error code

2025-10-29T09:57:06Z

commit 1ed0ff0000999561feee336c289252faf2502a7e parent 00bae5c394c9b1d31586ffa1ee2f77a3edcc6f0a Author: Roberto E. Vargas Caballero Date: Wed, 29 Oct 2025 10:57:06 +0100 tests: Set error code Having a correct error code when make tests is executed helps to detect when a test failed.

tests/make: Reduce the timeout for 0102-signal

2025-10-28T20:21:11Z

commit 00bae5c394c9b1d31586ffa1ee2f77a3edcc6f0a parent 65e5c373ca2fa947a604a1e0ea35843c8c104658 Author: Roberto E. Vargas Caballero Date: Tue, 28 Oct 2025 21:21:11 +0100 tests/make: Reduce the timeout for 0102-signal This timeout never had to be changed, because changing it was only hiding the actual problem, which was a race condition due to the incorrect signal handling.

make: Use a local signal variable

2025-10-28T20:12:08Z

commit 65e5c373ca2fa947a604a1e0ea35843c8c104658 parent 93e07bb34a825734d760971f811ac43868fd8c48 Author: Roberto E. Vargas Caballero Date: Tue, 28 Oct 2025 21:12:08 +0100 make: Use a local signal variable As the global stop variable is used in cleanup() we can be caught by one signal when we enter in cleanup but stop can be pointing to a different signal at the end of that function.

scc-make: Avoid signal race conditions

2025-10-28T20:10:31Z

commit 93e07bb34a825734d760971f811ac43868fd8c48 parent e19109cd35b8d64480c74389b8faaedf5af9b0ed Author: Roberto E. Vargas Caballero Date: Tue, 28 Oct 2025 21:10:31 +0100 scc-make: Avoid signal race conditions Expecting ait to be interrupted by a signal is too risky and it is very easy to be caught by race conditions. The best option is to kill in the snal handler and lock signals until we are ready to deal with them.

tests/make: Increment timeout for 0102-signal

2025-10-17T07:40:03Z

commit e19109cd35b8d64480c74389b8faaedf5af9b0ed parent 6d9a517cf4c6b5133030b9744531340d22cef719 Author: Roberto E. Vargas Caballero Date: Fri, 17 Oct 2025 09:40:03 +0200 tests/make: Increment timeout for 0102-signal This tests shows to be a bit flaky. Increment the time can help a bit.

build: Modify include path

2025-10-16T11:22:18Z

commit 6d9a517cf4c6b5133030b9744531340d22cef719 parent 6c3c992dc8f759f8c4052a4fb0d47d7681282664 Author: Roberto E. Vargas Caballero Date: Thu, 16 Oct 2025 13:22:18 +0200 build: Modify include path

Merge remote-tracking branch 'origin/master'

2025-08-05T12:00:25Z

commit 6c3c992dc8f759f8c4052a4fb0d47d7681282664 parent 23c081533234d616d3cc57e5794ee1b9f2a87b63 Author: Roberto E. Vargas Caballero Date: Tue, 5 Aug 2025 14:00:25 +0200 Merge remote-tracking branch 'origin/master'

make: Make cleanup void

2025-08-05T11:53:18Z

commit 23c081533234d616d3cc57e5794ee1b9f2a87b63 parent 7d92f9f22004c00fca2fe52a183b8d8e1a1bca6f Author: Roberto E. Vargas Caballero Date: Tue, 5 Aug 2025 13:53:18 +0200 make: Make cleanup void It does not return any value ever, so it does not make sense to have it like int, and it created problems porting it to plan9.

libc/malloc: Unify malloc and realloc

2025-06-17T18:16:23Z

commit 2a2e57ae043b6d78a5dbd50761971d6508540729 parent 2b33af36beceaf67507ab881f113d68a371a4f13 Author: Roberto E. Vargas Caballero Date: Tue, 17 Jun 2025 20:16:23 +0200 libc/malloc: Unify malloc and realloc realloc(p, 0) should behave like malloc(0), but we were not setting the same errno values, and we were returning NULL in both cases, that makes more difficult to differentiate the case when 0 is passed and the case when we don't have enough memory. Given that the traditional behaviour was always to return a pointer that could be freed (and even that was the original behaviour in scc) this commit restores that behaviour.

tests/make: Unset MAKEFLAGS

2025-06-17T19:04:13Z

commit 2b33af36beceaf67507ab881f113d68a371a4f13 parent ac449d18a971e3e452e3f47cbd2af90b6a55ea18 Author: Roberto E. Vargas Caballero Date: Tue, 17 Jun 2025 21:04:13 +0200 tests/make: Unset MAKEFLAGS Some make implementations set MAKEFLAGS to non portable values creating some problems when running the tests when we move from the driven make to the make under test.

scc-make: parseargv(): properly handle the -- argument, add test.

2025-06-13T20:31:14Z

commit ac449d18a971e3e452e3f47cbd2af90b6a55ea18 parent 30255fb57a7b5abf021e417aa34f380ae92bdbbe Author: Z. Gilboa Date: Fri, 13 Jun 2025 20:31:14 +0000 scc-make: parseargv(): properly handle the -- argument, add test.

.gitignore: add missing entry for /src/cmd/scc-dump.

2025-06-14T07:42:49Z

commit 30255fb57a7b5abf021e417aa34f380ae92bdbbe parent 63c66208e583d8538351428cb2d5bdf0aabe3e26 Author: Z. Gilboa Date: Sat, 14 Jun 2025 07:42:49 +0000 .gitignore: add missing entry for /src/cmd/scc-dump.

driver/posix: Enable QBE for any arch/abi

2025-06-04T14:36:22Z

commit 63c66208e583d8538351428cb2d5bdf0aabe3e26 parent 1f8f494a8504aa99052554476d86bcd0d6a7a1ba Author: Roberto E. Vargas Caballero Date: Wed, 4 Jun 2025 16:36:22 +0200 driver/posix: Enable QBE for any arch/abi The validation for the correct combination of QBE, arch and abi is done when the qbe target selection is done, so we don't have to care about it when we select the scc backend used to generate the QBE IL.

driver/posix: Adjust qbe command line based in the target

2025-06-04T14:31:49Z

commit 1f8f494a8504aa99052554476d86bcd0d6a7a1ba parent d853268357f02d13ca7377e2391b8ebb6290daad Author: Roberto E. Vargas Caballero Date: Wed, 4 Jun 2025 16:31:49 +0200 driver/posix: Adjust qbe command line based in the target Qbe supports multiple ABIs and they are selected using the command line flag -t.

cc2: Add support for qbe riscv64

2025-06-04T14:08:26Z

commit d853268357f02d13ca7377e2391b8ebb6290daad parent faf643c9afc0a74b276f24a8fdb4ba8552e351b3 Author: Roberto E. Vargas Caballero Date: Wed, 4 Jun 2025 16:08:26 +0200 cc2: Add support for qbe riscv64

cc2/arm64: Removed unused peep.c

2025-06-04T14:08:07Z

commit faf643c9afc0a74b276f24a8fdb4ba8552e351b3 parent 06266207ce41e9fa1b7c2232b0a0243481287fd6 Author: Roberto E. Vargas Caballero Date: Wed, 4 Jun 2025 16:08:07 +0200 cc2/arm64: Removed unused peep.c This file was only a placeholder for the times when we built a skeleton for arm64 without using qbe. As we don't build that skeleton anymore, it does not make sense to keep this placeholder anymore in the repository.

cc1: Add support for riscv64

2025-06-04T14:06:01Z

commit 06266207ce41e9fa1b7c2232b0a0243481287fd6 parent 7d92f9f22004c00fca2fe52a183b8d8e1a1bca6f Author: Roberto E. Vargas Caballero Date: Wed, 4 Jun 2025 16:06:01 +0200 cc1: Add support for riscv64

build: Remove redundant make build

2025-05-27T18:19:30Z

commit 7d92f9f22004c00fca2fe52a183b8d8e1a1bca6f parent 6182ec6e314de95182742064f5bb2d80abf11c8c Author: Roberto E. Vargas Caballero Date: Tue, 27 May 2025 20:19:30 +0200 build: Remove redundant make build Scc-make was already built before this rule began its execution.

libc/vfprintf: Adjust variable types

2025-05-24T04:54:25Z

commit 6182ec6e314de95182742064f5bb2d80abf11c8c parent 9c55c9b38fb53863c83f8d59ea6241682b2ea463 Author: Roberto E. Vargas Caballero Date: Sat, 24 May 2025 06:54:25 +0200 libc/vfprintf: Adjust variable types

doc: Clarify how to use musl

2025-05-13T08:14:24Z

commit 9c55c9b38fb53863c83f8d59ea6241682b2ea463 parent 5900420c8566844b1e379dc62adb0467142f56d6 Author: Roberto E. Vargas Caballero Date: Tue, 13 May 2025 10:14:24 +0200 doc: Clarify how to use musl