LADSoft
omake in linux
I extracted omake from the orangec source tree (copied all needed source files until the compiler stopped complaining) and added them into a RedPanda DevCpp project. I added -DHAVE_UNISTD_H and -pthread. It compiled fine with a few warnings. But the warning by the linker seems to need attention.
/usr/bin/ld: ./omake/os.o: in function `OS::JobInit()':
os.cpp:(.text+0x98a): warning: the use of `tempnam' is dangerous, better use `mkstemp'
/usr/bin/ld: ./util/Utils.o: in function `Utils::TempName(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&)':
Utils.cpp:(.text+0xd24): warning: the use of `tmpnam' is dangerous, better use `mkstemp'
Anyway, I have a make binary (I named my omake binary simply make).
The problem is this omake doesn't work with any makefile. It always crashes with this error:
terminate called after throwing an instance of 'std::system_error'
what(): Bad file descriptor
Aborted
It seems it's looking for make.cfg or unknown.cfg which is not available. On Windows, it works fine without any of these files.
Here is the result of strace:
$ strace ./make
execve("./make", ["./make"], 0x7ffc3c43d880 /* 45 vars */) = 0
brk(NULL) = 0x5e9c893ec000
mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f5adb720000
access("/etc/ld.so.preload", R_OK) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
newfstatat(3, "", {st_mode=S_IFREG|0644, st_size=78862, ...}, AT_EMPTY_PATH) = 0
mmap(NULL, 78862, PROT_READ, MAP_PRIVATE, 3, 0) = 0x7f5adb70c000
close(3) = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libstdc++.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
newfstatat(3, "", {st_mode=S_IFREG|0644, st_size=2190440, ...}, AT_EMPTY_PATH) = 0
mmap(NULL, 2201728, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7f5adb400000
mmap(0x7f5adb499000, 1052672, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x99000) = 0x7f5adb499000
mmap(0x7f5adb59a000, 454656, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x19a000) = 0x7f5adb59a000
mmap(0x7f5adb609000, 57344, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x209000) = 0x7f5adb609000
mmap(0x7f5adb617000, 10368, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x7f5adb617000
close(3) = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libgcc_s.so.1", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
newfstatat(3, "", {st_mode=S_IFREG|0644, st_size=125312, ...}, AT_EMPTY_PATH) = 0
mmap(NULL, 127688, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7f5adb6ec000
mmap(0x7f5adb6ef000, 94208, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x3000) = 0x7f5adb6ef000
mmap(0x7f5adb706000, 16384, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1a000) = 0x7f5adb706000
mmap(0x7f5adb70a000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1d000) = 0x7f5adb70a000
close(3) = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libc.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\20t\2\0\0\0\0\0"..., 832) = 832
pread64(3, "\6\0\0\0\4\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0"..., 784, 64) = 784
newfstatat(3, "", {st_mode=S_IFREG|0755, st_size=1922136, ...}, AT_EMPTY_PATH) = 0
pread64(3, "\6\0\0\0\4\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0"..., 784, 64) = 784
mmap(NULL, 1970000, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7f5adb21f000
mmap(0x7f5adb245000, 1396736, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x26000) = 0x7f5adb245000
mmap(0x7f5adb39a000, 339968, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x17b000) = 0x7f5adb39a000
mmap(0x7f5adb3ed000, 24576, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1ce000) = 0x7f5adb3ed000
mmap(0x7f5adb3f3000, 53072, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x7f5adb3f3000
close(3) = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libm.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
newfstatat(3, "", {st_mode=S_IFREG|0644, st_size=907784, ...}, AT_EMPTY_PATH) = 0
mmap(NULL, 909560, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7f5adb140000
mmap(0x7f5adb150000, 471040, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x10000) = 0x7f5adb150000
mmap(0x7f5adb1c3000, 368640, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x83000) = 0x7f5adb1c3000
mmap(0x7f5adb21d000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0xdc000) = 0x7f5adb21d000
close(3) = 0
mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f5adb6ea000
mmap(NULL, 12288, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f5adb6e7000
arch_prctl(ARCH_SET_FS, 0x7f5adb6e7740) = 0
set_tid_address(0x7f5adb6e7a10) = 3777
set_robust_list(0x7f5adb6e7a20, 24) = 0
rseq(0x7f5adb6e8060, 0x20, 0, 0x53053053) = 0
mprotect(0x7f5adb3ed000, 16384, PROT_READ) = 0
mprotect(0x7f5adb21d000, 4096, PROT_READ) = 0
mprotect(0x7f5adb70a000, 4096, PROT_READ) = 0
mprotect(0x7f5adb609000, 45056, PROT_READ) = 0
mprotect(0x5e9c6c8e2000, 4096, PROT_READ) = 0
mprotect(0x7f5adb758000, 8192, PROT_READ) = 0
prlimit64(0, RLIMIT_STACK, NULL, {rlim_cur=8192*1024, rlim_max=RLIM64_INFINITY}) = 0
munmap(0x7f5adb70c000, 78862) = 0
getrandom("\x69\x82\x0a\xe3\x27\x95\x24\x56", 8, GRND_NONBLOCK) = 8
brk(NULL) = 0x5e9c893ec000
brk(0x5e9c8940d000) = 0x5e9c8940d000
futex(0x7f5adb61773c, FUTEX_WAKE_PRIVATE, 2147483647) = 0
access("./make.cfg", F_OK) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "unknown.cfg", O_RDONLY) = -1 ENOENT (No such file or directory)
pipe2([3, 4], 0) = 0
write(-1, "1", 1) = -1 EBADF (Bad file descriptor)
futex(0x7f5adb70b1f0, FUTEX_WAKE_PRIVATE, 2147483647) = 0
write(2, "terminate called after throwing "..., 48terminate called after throwing an instance of ') = 48
write(2, "std::system_error", 17std::system_error) = 17
write(2, "'\n", 2'
) = 2
write(2, " what(): ", 11 what(): ) = 11
write(2, "Bad file descriptor", 19Bad file descriptor) = 19
write(2, "\n", 1
) = 1
rt_sigprocmask(SIG_UNBLOCK, [ABRT], NULL, 8) = 0
gettid() = 3777
getpid() = 3777
tgkill(3777, 3777, SIGABRT) = 0
--- SIGABRT {si_signo=SIGABRT, si_code=SI_TKILL, si_pid=3777, si_uid=1000} ---
+++ killed by SIGABRT +++
Aborted
Please move that to https://github.com/LADSoft/OMAKE, which should also make it easy to build without manual deletion of files.
I am utterly unsurprised that omake doesn't function on linux, I'm more surprised that it's a bad file descriptor.
A lot of the "OS" wrapper libraries that are set up for os-independence aren't implemented on linux at the moment, after the next release I wanted to go after tackling that so that there's some modicum of "linux" support (even if it's not yet the compiler supporting compiling TO linux, just on), outside of that there shouldn't be much of anything if at all being a blocker. If you want, I'd try modifying os.cpp such that everything supported on windows is supported on Linux and seeing if that fixes anything....
As for the make.cfg issue, try changing ToolChain.cpp in the util, the
access call to have "F_OK", I think on Linux the F_OK call is something
different than on windows and that is the likely cause there...
Also, renaming it when you should have omake.cfg will cause it to hunt for
make.cfg for the check, so that's another weird thing...
On Mon, Sep 9, 2024 at 1:54 PM Simon Sobisch @.***> wrote:
Please move that to https://github.com/LADSoft/OMAKE, which should also make it easy to build without manual deletion of files.
— Reply to this email directly, view it on GitHub https://github.com/LADSoft/OrangeC/issues/1044#issuecomment-2338732386, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSPBTXQXBASAUHW27CHP23ZVXOGZAVCNFSM6AAAAABN5BGMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMZYG4ZTEMZYGY . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Please move that to https://github.com/LADSoft/OMAKE, which should also make it easy to build without manual deletion of files.
This repository is not synced with the latest changes. Getting your build system to work on Linux is also a hassle. This is why I decided to use my own build system (a RedPanda DevCpp project).
@chuggafan I changed line 110 in ToolChain.cpp from if (access(configName.c_str(), 0) != 0) to if (access(configName.c_str(), 0) != F_OK) but it doesn't help. ~~I'm willing to test your changes. Please tell me where could I clone your code.~~
Note: I build omake completely decoupled from orange c. ToolChain::StandardToolStartup makes a reference to the ORANGEC environment variable. It seems omake is tied to orange c? I don't have orange c. Building omake alone in linux is easy. But the whole orange c suite is another story.
Please move that to https://github.com/LADSoft/OMAKE, which should also make it easy to build without manual deletion of files.
This repository is not synced with the latest changes.
Right. That repo's state is considered to be stable and has an issue tracker specific for omake, while this repo's master is both focused on the whole suite and on "latest testing" state.
Note: I build omake completely decoupled from orange c.
ToolChain::StandardToolStartupmakes a reference to theORANGECenvironment variable. It seems omake is tied to orange c?
That's a reasonable thing for the defaults - the builtin rules need an internal define for CC/CXX/AS (= if not explicit specified in the makefile) and those use the OrangeC tools; their fullpath is resolved from the ORANGEC environment variable and if not already set this is done internally, relative from omakes path.
I've done some investigation, it seems to be something up with my jobserver implementation, I only did a preliminary investigation and I'll do a further one over the weekend but it seems like my "quick fix" is actually "wtf am I doing on Unix again?" Probably from not initializing job checks, we'll see...
On Tue, Sep 10, 2024 at 3:37 AM Simon Sobisch @.***> wrote:
Please move that to https://github.com/LADSoft/OMAKE, which should also make it easy to build without manual deletion of files.
This repository is not synced with the latest changes.
Right. That repo's state is considered to be stable and has an issue tracker specific for omake, while this repo's master is both focused on the whole suite and on "latest testing" state.
Note: I build omake completely decoupled from orange c. ToolChain::StandardToolStartup makes a reference to the ORANGEC environment variable. It seems omake is tied to orange c?
That's a reasonable thing for the defaults - the builtin rules need an internal define for CC/CXX/AS (= if not explicit specified in the makefile) and those use the OrangeC tools; their fullpath is resolved from the ORANGEC environment variable and if not already set this is done internally, relative from omakes path.
— Reply to this email directly, view it on GitHub https://github.com/LADSoft/OrangeC/issues/1044#issuecomment-2339897300, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSPBTSQ6LASQOVB7KW6JQ3ZV2OTVAVCNFSM6AAAAABN5BGMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMZZHA4TOMZQGA . You are receiving this because you were mentioned.Message ID: @.***>
Also, renaming it when you should have omake.cfg will cause it to hunt for make.cfg for the check, so that's another weird thing...
@chuggafan Please have a look at your ToolChain::StandardToolStartup method:
https://github.com/LADSoft/OrangeC/blob/master/src/util/ToolChain.cpp#L99
The name of the .cfg file will be the name of the make binary + .cfg. I name my omake binary make, so the .cfg file is make.cfg. Nothing weird. unknown.cfg should be somewhere in the code. Maybe when it can't determine the current make binary name, it will use unknown. I don't know for sure. I'm not dive into the code yet.
I've done some investigation, it seems to be something up with my jobserver implementation, I only did a preliminary investigation and I'll do a further one over the weekend but it seems like my "quick fix" is actually "wtf am I doing on Unix again?" Probably from not initializing job checks, we'll see...
The problem is I only use just one make job (doesn't specify -j at all). It seems you are going the wrong direction. Compiling it with debug and running it in gdb will reveal interesting things. Everything points to the BUILTINS in MakeMain.cpp and thus the .cfg file and the ToolChain::StandardToolStartup method.
The install/package step should rename the file - but as an external build system is used here this is possiblity already done - if there are any options to override the binary names in our Makefiles in the first place...
I managed to compile the whole orange c suite in linux. This is the command I used:
make COMPILER="gcc-linux"
But I don't know what to do after that. How to install it so I could set the ORANGEC environment variable?
Btw, I noticed that the binaries have .exe extensions even though it's Linux. I checked using file and they are indeed Linux ELF binaries.
Frankly speaking, I'm more interested in omake as a standalone tool. If I need the whole orange c suite to use omake then it's a lot less appealing to me.
Please move that to https://github.com/LADSoft/OMAKE, which should also make it easy to build without manual deletion of files.
This repository is not synced with the latest changes.
Right. That repo's state is considered to be stable and has an issue tracker specific for omake, while this repo's master is both focused on the whole suite and on "latest testing" state.
The omake in this repository will not even compile with make COMPILER="gcc-linux". It's full of compilation errors.
Mr. ghost (#1038) is absolutely right. Please run static analyzers on your code. Or at least try compiling using latest GCC and Clang with the highest diagnostics. Goodbye.
Thanks for the note, but the issue should be kept open as it is unresolved (please opt out of subscription instead if you don't want to receive any updates).
Thanks for the note, but the issue should be kept open as it is unresolved (please opt out of subscription instead if you don't want to receive any updates).
OK.
I'm working on this issue now that it's the weekend, seems like it's a multilayered issue with omake, the majority of it seems (to me) to be individual little bits that I got wrong from not testing on Linux before getting into the real meat of issues (i.e. semaphores being broken, the jobserver pipe being broken, etc.) from making bad assumptions. I'm now debugging why the VariableContainer isn't getting linux environment variables, which will be the next steps to being able to say: "Hey, the shell is /bin/sh, please stop this annoying stuff"
On Thu, Sep 12, 2024 at 2:41 AM mh466lfa @.***> wrote:
Thanks for the note, but the issue should be kept open as it is unresolved (please opt out of subscription instead if you don't want to receive any updates).
OK.
— Reply to this email directly, view it on GitHub https://github.com/LADSoft/OrangeC/issues/1044#issuecomment-2345395539, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSPBTTH55357NUZV27WGM3ZWEZRTAVCNFSM6AAAAABN5BGMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGNBVGM4TKNJTHE . You are receiving this because you were mentioned.Message ID: @.***>
Please consider MSYS and possibly cygwin - using a test for the shell may break some parts if that's for finding the OS...
You do know about the OS environment variable, right?
I do know of the OS environment variable being a thing on Windows, however
I also am aware that on Linux (specifically my WSL debian, Ubuntu, work's
RHEL 9.4 and 8.10 servers as well) installation SHELL is a defined
variable (and MSYS also does this).
Windows doesn't do this by default, but I'm taking the already (mediocre) circuitry and just shorting it so that if it detects SHELL, it looks for "sh" and sets the internal mechanisms to say "This is a (sh)ell system (as I've found that pretty much every shell ends in sh, dash, bash, sh, zsh, fish) and there's only a few I've found that don't (cancer (named because it's terminal)).
Ideally we'd have separate compiled instances for windows and linux, which is already somewhat of a thing, but needs more mechanisms to make that truly happen outside of weird wine executions.
All in all this change is a bit more complex than I first realized and so
will probably take a few days to nail down each individual thing so that
omake is even passable on linux, then I can strip out every single SHELL=
thing in the makefiles so that we're not erroneously overriding that in WSL
where cmd.exe is actually available and breaks the builds.
On Sat, Sep 14, 2024 at 3:00 AM Simon Sobisch @.***> wrote:
Please consider MSYS and possibly cygwin - using a test for the shell may break some parts if that's for finding the OS...
You do know about the OS environment variable, right?
— Reply to this email directly, view it on GitHub https://github.com/LADSoft/OrangeC/issues/1044#issuecomment-2350882903, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSPBTTBSDZLZIUCLDXKYSDZWPNHNAVCNFSM6AAAAABN5BGMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGNJQHA4DEOJQGM . You are receiving this because you were mentioned.Message ID: @.***>
I do miss something. Within the compile we have #ifdef _WIN32, for the build system we can explicitly set variables during make as well...
@chuggafan thank you for looking into this.
FWIW I've been trying to get away from checking the _WIN32 define and I made a define TARGET_OS_WINDOWS... the intention was when that isn't set it is for linux/unix but I don't suppose there is any reason we couldn't additionally add TARGET_OS_LINUX if that seems reasonable....
I mean, I understand the TARGET_OS_WINDOWS vs TARGET_OS_LINUX distinction, but fundamentally I think it's mostly just a "smart renaming" of _WIN32, and unix/linux/whatever gcc and clang are doing since realistically they're the only relevant players outside of EDG compilers, and even those define unix when targeting unix-like platforms https://godbolt.org/z/Yj8bvbc5a I'm more than fine continuing with the smart renaming aspect, it's more so just that there's already existing mechanisms that are intended for exactly this kind of "check what platform we're on" stuff, I mostly just want to wrap all of that into a little library like the OS libraries and get on with that...
On Sat, Sep 14, 2024 at 1:39 PM David Lindauer @.***> wrote:
@chuggafan https://github.com/chuggafan thank you for looking into this.
FWIW I've been trying to get away from checking the _WIN32 define and I made a define TARGET_OS_WINDOWS... the intention was when that isn't set it is for linux/unix but I don't suppose there is any reason we couldn't additionally add TARGET_OS_LINUX if that seems reasonable....
— Reply to this email directly, view it on GitHub https://github.com/LADSoft/OrangeC/issues/1044#issuecomment-2351080104, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSPBTWLOG7RON6GDXKDKC3ZWRYD5AVCNFSM6AAAAABN5BGMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGNJRGA4DAMJQGQ . You are receiving this because you were mentioned.Message ID: @.***>
Whelp, the issue I ran into was simpler than I thought: environ is a thing in POSIX as well as WINDOWS, just does it a little differently, so I modified that. OMAKE STILL does not run properly on linux, but it's no longer crashing and (should) define the shell variable. It's still bugging out at dealing with pretty much everything inside of OS, it's just we now get environment variables 🎉 The places where this is still an os.cpp issue are process spawning (I don't even know where to begin, on UNIX there's 8 different ways to do it and it's unclear which is the "right" way nowadays, whereas on windows there's a grand total of 2), GetFullPath, TerminateAll, and everything to do with file times. Work does also need to be done on using the default shell $(SHELL) everywhere possible, especially since on windows omake is setting that to cmd.exe by default (I checked), but that can be done at a later date. I'll push my changes to a separate branch, and notify you about it to make a PR sometime today/tomorrow depending on if I feel like tackling at least some of the os.cpp changes that need to go through for proper omake functioning on linux.
On Sat, Sep 14, 2024 at 2:02 PM chugga fan @.***> wrote:
I mean, I understand the TARGET_OS_WINDOWS vs TARGET_OS_LINUX distinction, but fundamentally I think it's mostly just a "smart renaming" of _WIN32, and unix/linux/whatever gcc and clang are doing since realistically they're the only relevant players outside of EDG compilers, and even those define unix when targeting unix-like platforms https://godbolt.org/z/Yj8bvbc5a I'm more than fine continuing with the smart renaming aspect, it's more so just that there's already existing mechanisms that are intended for exactly this kind of "check what platform we're on" stuff, I mostly just want to wrap all of that into a little library like the OS libraries and get on with that...
On Sat, Sep 14, 2024 at 1:39 PM David Lindauer @.***> wrote:
@chuggafan https://github.com/chuggafan thank you for looking into this.
FWIW I've been trying to get away from checking the _WIN32 define and I made a define TARGET_OS_WINDOWS... the intention was when that isn't set it is for linux/unix but I don't suppose there is any reason we couldn't additionally add TARGET_OS_LINUX if that seems reasonable....
— Reply to this email directly, view it on GitHub https://github.com/LADSoft/OrangeC/issues/1044#issuecomment-2351080104, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSPBTWLOG7RON6GDXKDKC3ZWRYD5AVCNFSM6AAAAABN5BGMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGNJRGA4DAMJQGQ . You are receiving this because you were mentioned.Message ID: @.***>
Look at this for a reference on POSIX system (how to spawn new processes, environment variables, blah blah...):
https://github.com/rmyorston/pdpmake
Btw, I suggest using POSIX instead of Linux. You don't want your software to run on other *nixes like FreeBSD?
Your code is GPL-ed, and since the GPL can absorb code written in weaker licenses, I don't see there are any reasons to not look into the code of GNU make itself or BSD make and even copy it.
bmake is here: https://www.crufty.net/help/sjg/bmake.html
uname woes on MSYS2: https://codeberg.org/schilytools/schilytools/issues/60
Note: CDDL is not compatible with GPL.
Please consider MSYS and possibly cygwin - using a test for the shell may break some parts if that's for finding the OS...
You do know about the
OSenvironment variable, right?
Orange C failed to link on Cygwin and FreeBSD. Compilation is fine, but linking is not. Why?
And as @mh466lfa has noted, the executables have .exe extensions even on Linux and FreeBSD.