Building ACK with MSYS2

[KristaMHill]

Hello folks, after following up with the ACK discussions and looking about I'm noting the following:

o With Win, the MSYS tools appears to be suggested
o The ninja build system is also used, which is easy to install using pacman
o In using the MSYS2 UCRT64 environment, soon after many .py files appear to be loaded, an error is encountered:
o I next tried the MSYS2 MINGW64 environment and had a similar result. After the make command, again many .py files appear to be loaded, but this time, much slower. A similar error appeared:

In looking about it appears that rm doesn't support the "--link -s" options

==========================
The following is snipped from the MSYS2 UCRT64 environment

[prior output removed]
loading plat/linuxppc/libsys/build.py
loading examples/build.py
[1/7129] RULE h/+local
FAILED: [code=1] .obj/h/+local/local.h
rm -rf .obj/h/+local/sandbox&&python3 build/_sandbox.py --link -s .obj/h/+local/sandbox&&(cd .obj/h/
+local/sandbox && mkdir -p .obj/h/+local )&&(cd .obj/h/+local/sandbox && echo '#define VERSION 3' >
.obj/h/+local/local.h )&&(cd .obj/h/+local/sandbox && echo '#define ACKM "pc86"' >> .obj/h/+local/lo
cal.h )&&(cd .obj/h/+local/sandbox && echo '#define BIGMACHINE 1' >> .obj/h/+local/local.h )&&(cd .o
bj/h/+local/sandbox && echo '#define SYS_5' >> .obj/h/+local/local.h )&&python3 build/_sandbox.py --
export -s .obj/h/+local/sandbox .obj/h/+local/local.h
rm: unknown option -- link
Try 'rm --help' for more information.

[Much further below there was an an error summary]

ninja: build stopped: subcommand failed.
make: *** [build/ab.mk:126: +all] Error 1

==================

The following is snipped from the MSYS2 MINGW64 environment build

[prior output removed]
loading plat/pc86/emu/build.py
/bin/sh: line 1: pkg-config: command not found
[1/9927] CHEADERS modules/src/data/+data_hdr
FAILED: [code=1] .obj/modules/src/data/+data_hdr/array.h .obj/modules/src/data/+data_hdr/astring.h .ob
j/modules/src/data/+data_hdr/warnings.h .obj/modules/src/data/+data_hdr/imap.h .obj/modules/src/data/+
data_hdr/pmap.h .obj/modules/src/data/+data_hdr/smap.h .obj/modules/src/data/+data_hdr/stringlist.h
rm -rf .obj/modules/src/data/+data_hdr/sandbox&&python3 build/_sandbox.py --link -s .obj/modules/src/d
ata/+data_hdr/sandbox modules/src/data/imap.h modules/src/data/smap.h modules/src/data/astring.h modul
es/src/data/array.h modules/src/data/stringlist.h modules/src/data/warnings.h modules/src/data/pmap.h&
&(cd .obj/modules/src/data/+data_hdr/sandbox && mkdir -p .obj/modules/src/data/+data_hdr/. )&&(cd .obj
/modules/src/data/+data_hdr/sandbox && cp modules/src/data/array.h .obj/modules/src/data/+data_hdr/./a
rray.h )&&(cd .obj/modules/src/data/+data_hdr/sandbox && cp modules/src/data/astring.h .obj/modules/sr
c/data/+data_hdr/./astring.h )&&(cd .obj/modules/src/data/+data_hdr/sandbox && cp modules/src/data/war
nings.h .obj/modules/src/data/+data_hdr/./warnings.h )&&(cd .obj/modules/src/data/+data_hdr/sandbox &&
cp modules/src/data/imap.h .obj/modules/src/data/+data_hdr/./imap.h )&&(cd .obj/modules/src/data/+dat
a_hdr/sandbox && cp modules/src/data/pmap.h .obj/modules/src/data/+data_hdr/./pmap.h )&&(cd .obj/modul
es/src/data/+data_hdr/sandbox && cp modules/src/data/smap.h .obj/modules/src/data/+data_hdr/./smap.h )
&&(cd .obj/modules/src/data/+data_hdr/sandbox && cp modules/src/data/stringlist.h .obj/modules/src/dat
a/+data_hdr/./stringlist.h )&&python3 build/_sandbox.py --export -s .obj/modules/src/data/+data_hdr/sa
ndbox .obj/modules/src/data/+data_hdr/./array.h .obj/modules/src/data/+data_hdr/./astring.h .obj/modul
es/src/data/+data_hdr/./warnings.h .obj/modules/src/data/+data_hdr/./imap.h .obj/modules/src/data/+dat
a_hdr/./pmap.h .obj/modules/src/data/+data_hdr/./smap.h .obj/modules/src/data/+data_hdr/./stringlist.h
rm: unknown option -- link
Try 'rm --help' for more information.

Then much further down, there was an error summary:

ninja: build stopped: subcommand failed.
make: *** [build/ab.mk:126: +all] Error 1

================
Your advice in this matter will be appreciated.
Krista

[KristaMHill]

Just to add, I looked in ack > plat >pc86 > emu > build.py and found that line 1 in that file has:

from build.c import hostclibrary, hostcprogram

This is strange as the error is reported from /bin/sh, rather than Python. Also at the top of build.py, would a "Sh-bang" be expected to refer to Python?

loading plat/pc86/emu/build.py
/bin/sh: line 1: pkg-config: command not found

Aty this point I'm just guessing...
Krista

[davidgiven]

I see this in the command line:

rm -rf .obj/h/+local/sandbox&&python3 build/_sandbox.py --link -s ...

The && is supposed to separate commands. I suspect your shell isn't recognising it without whitespace around it. If so, that's a build system bug, not your fault, and I'll take a look, but I haven't seen that before. It's also possible it's deleted build/_sandbox.py, so you might want to check for that.

Also, Windows doesn't like the sandbox script I'm using. The known good command line invocation (from the .github/workflows/ccpp.yaml file which is used for the nightly builds) is:

make LDFLAGS="-s -static" ack-setup.exe AB_SANDBOX=no

I forget why the -s -static is there, though. That file's also a good place to check your dependencies.

The ACK, by the way, is now much easier to build than it was in the 5.5 days, believe it or not...

[KristaMHill]

Just looked in the attempted MSYS2 UCRT64 environment build as well as the attempted MSYS2 MINGW build and did find build/_sandbox.py in there.

Regarding build the ACK, I'll take your word that it's easier than version 5.5. In any regard, thanks so much for your efforts, they are impressive and appreciated. I'm patient and know that it's one step at a time.

Just unpacked a fresh download of the ACK build files and with MSYS2 UCRT64 tried the make command you just suggested. Well, the build still fails in about the same place, This time there is less space between announced failures.

[Many lines removed]
loading plat/linuxppc/libsys/build.py
loading examples/build.py
[1/7129] HOSTCC util/LLgen/+LLgen/util/LLgen/src/LLgen.c
FAILED: [code=1] .obj/util/LLgen/+LLgen/util/LLgen/src/LLgen.c/LLgen.o
mkdir -p .obj/util/LLgen/+LLgen/util/LLgen/src/LLgen.c&&gcc -c -o .obj/util/LLgen/+LLgen/util/LLgen/
src/LLgen.c/LLgen.o util/LLgen/src/LLgen.c -g -Os -Werror=return-type -Werror=implicit-function-decl
aration -Werror=strict-prototypes -DUNREACHABLE_CODE='__builtin_unreachable()' -DNORETURN=_Noreturn
-DLIBDIR="util/LLgen/lib" -DNON_CORRECTING -Iutil/LLgen/src
mkdir: unknown option -- c
Try 'mkdir --help' for more information.

[Some other errors indicate that a file already exists: ]

[3/7129] CHEADERS modules/src/system/+system_hdr
FAILED: [code=1] .obj/modules/src/system/+system_hdr/system.h
mkdir -p .obj/modules/src/system/+system_hdr&&cp modules/src/system/system.h .obj/modules/src/system
/+system_hdr/system.h
mkdir: cannot create directory ΓÇÿmodules/src/system/system.hΓÇÖ: File exists

'Best,
Krista

[KristaMHill]

Just to add, when I look at the files at the top of the build directory, I see the following. Again, I'm just guessing...
Krista

$ ls
'#define ACKM "pc86"' '<' Makefile build.py fast mach
'#define ACK_PATH "share' '>' NEW compile_commands.json fcc man
'#define BIGMACHINE 1' '>>' README doc h modules
'#define EM_DIR "c:' Action TODO emtest include plat
'#define SYS_5' CHANGES bin etc lang tests
'#define VERSION 3' Copyright build examples lib util

[davidgiven]

Could you try the version at #364 and let me know if that helps any? It looks like you'll need to completely wipe your working tree, as it's likely files are corrupted.

[KristaMHill]

Thanks, I'll take a look. Krista

…

________________________________ From: David Given ***@***.***> Sent: Thursday, April 23, 2026 5:57 PM To: davidgiven/ack ***@***.***> Cc: Hill, Krista ***@***.***>; Author ***@***.***> Subject: [External-email]: Re: [davidgiven/ack] Building ACK with MSYS2 (Discussion #363) **Caution: This email was sent from outside the university. Please exercise caution when clicking links, opening attachments, or replying. When in doubt, contact the helpdesk.** Could you try the version at #364<#364> and let me know if that helps any? It looks like you'll need to completely wipe your working tree, as it's likely files are corrupted. — Reply to this email directly, view it on GitHub<#363?email_source=notifications&email_token=AAYSHRWKKVQDGLXYDD2N6KD4XKGV7A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNRWHEYDSNBWUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVRTG633UMVZF6Y3MNFRWW#discussioncomment-16690946>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAYSHRRHL7QZ7EU4OAVZPNT4XKGV7AVCNFSM6AAAAACYC54QZWVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNRZGA4TINQ>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS<https://github.com/notifications/mobile/ios/AAYSHRSP6BGPB7Q3N5VSCPD4XKGV7A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNRWHEYDSNBWUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVJTG633UMVZF62LPOM> and Android<https://github.com/notifications/mobile/android/AAYSHRQUDFANZVA7432KM2L4XKGV7A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNRWHEYDSNBWUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVZTG633UMVZF6YLOMRZG62LE>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[KristaMHill]

I'm still new at this so, here goes. Using MSYS2 UCRT64 I cloned the ACK repo, then from within that I used gh to get pull 364. The .py files loaded as before, then a similar error. This time I can see whitespace about the && characters, which is good. I have no idea why an error is being caught with the rm command.

Also, as before, it's ninja that provides the summary:

ninja: build stopped: subcommand failed.
make: *** [build/ab.mk:126: +all] Error 1

Thanks,
Krista

[Many lines removed]
loading plat/linuxppc/libsys/build.py
loading examples/build.py
[1/7129] RULE h/+em_path
FAILED: [code=1] .obj/h/+em_path/em_path.h
rm -rf .obj/h/+em_path/sandbox && python3 build/_sandbox.py --link -s .obj/h/+em_path/sandbox && (cd
.obj/h/+em_path/sandbox && mkdir -p .obj/h/+em_path ) && (cd .obj/h/+em_path/sandbox && echo '#defi
ne EM_DIR "c:/ack"' > .obj/h/+em_path/em_path.h ) && (cd .obj/h/+em_path/sandbox && echo '#define AC
K_PATH "share/ack/descr"' >> .obj/h/+em_path/em_path.h ) && python3 build/_sandbox.py --export -s .o
bj/h/+em_path/sandbox .obj/h/+em_path/em_path.h
rm: unknown option -- link
Try 'rm --help' for more information.
[2/7129] CHEADERS modules/src/system/+system_hdr
FAILED: [code=1] .obj/modules/src/system/+system_hdr/system.h
rm -rf .obj/modules/src/system/+system_hdr/sandbox && python3 build/_sandbox.py --link -s .obj/modul
es/src/system/+system_hdr/sandbox modules/src/system/system.h && (cd .obj/modules/src/system/+system
_hdr/sandbox && mkdir -p .obj/modules/src/system/+system_hdr ) && (cd .obj/modules/src/system/+syste
m_hdr/sandbox && cp modules/src/system/system.h .obj/modules/src/system/+system_hdr/system.h ) && py
thon3 build/_sandbox.py --export -s .obj/modules/src/system/+system_hdr/sandbox .obj/modules/src/sys
tem/+system_hdr/system.h
rm: unknown option -- link

[davidgiven]

Wait, I think I know what's going on! I bet you have a Windows-specific ninja installation installed, via chocolatey or something. That's appearing on your path and MSYS is finding it before the MSYS-specific one. The Windows ninja is executing commands via cmd.exe rather than sh, so the && shell syntax isn't being parsed and instead the entire command line is going to rm which of course is completely confused about it.

Try which ninja and see what it says.

[KristaMHill]

What you described seems right to me... if I call gcc from the command interpreter, then the MSYS2 version of gcc replies, so something is going on... C:\Users\Hill>gcc --version gcc (Rev14, Built by MSYS2 project) 15.2.0 [Copyright notice snipped] I next tried MSYS2 with ninja [PATH] UCRT64 ~ $ ninja --version 1.13.2 [PATH] UCRT64 ~ $ which ninja /ucrt64/bin/ninja And I next tried the command interpreter C:\Users\Hill>ninja ninja: error: loading 'build.ninja': The system cannot find the file specified. I'm thinking that something else is invoking ninja. In looking at the PATH in the environment variable editor I found C:\msys64\ucrt64\bin in there, which I removed. So now I cannot call gcc from the command interpreter. Just recloned ACK, grabbed pull 364, and tried to build. Again the .py files were loaded and then the FAILED error was encountered and eventually ninja produces the error summary. Thanks again, so far. I'll look about for anything... Krista ===================== [lines removed] loading plat/linuxppc/libsys/build.py loading examples/build.py [1/7129] CHEADERS modules/src/alloc/+alloc_hdr FAILED: [code=1] .obj/modules/src/alloc/+alloc_hdr/alloc.h rm -rf .obj/modules/src/alloc/+alloc_hdr/sandbox && python3 build/_sandbox.py --link -s .obj/modules /src/alloc/+alloc_hdr/sandbox modules/src/alloc/alloc.h && (cd .obj/modules/src/alloc/+alloc_hdr/san dbox && mkdir -p .obj/modules/src/alloc/+alloc_hdr ) && (cd .obj/modules/src/alloc/+alloc_hdr/sandbo x && cp modules/src/alloc/alloc.h .obj/modules/src/alloc/+alloc_hdr/alloc.h ) && python3 build/_sand box.py --export -s .obj/modules/src/alloc/+alloc_hdr/sandbox .obj/modules/src/alloc/+alloc_hdr/alloc .h rm: unknown option -- link Try 'rm --help' for more information. [lines removed] ninja: build stopped: subcommand failed. make: *** [build/ab.mk:126: +all] Error 1

…

________________________________ From: David Given ***@***.***> Sent: Friday, April 24, 2026 6:07 PM To: davidgiven/ack ***@***.***> Cc: Hill, Krista ***@***.***>; Author ***@***.***> Subject: [External-email]: Re: [davidgiven/ack] Building ACK with MSYS2 (Discussion #363) **Caution: This email was sent from outside the university. Please exercise caution when clicking links, opening attachments, or replying. When in doubt, contact the helpdesk.** Wait, I think I know what's going on! I bet you have a Windows-specific ninja installation installed, via chocolatey or something. That's appearing on your path and MSYS is finding it before the MSYS-specific one. The Windows ninja is executing commands via cmd.exe rather than sh, so the && shell syntax isn't being parsed and instead the entire command line is going to rm which of course is completely confused about it. Try which ninja and see what it says. — Reply to this email directly, view it on GitHub<#363 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAYSHRVOQWAO4UEK2ZRPFT34XPQRBAVCNFSM6AAAAACYC54QZWVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNZQG4ZDCMI>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS<https://github.com/notifications/mobile/ios/AAYSHRTK6OVKD7W23I5EMFT4XPQRBA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNRXGA3TEMJRUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVJTG633UMVZF62LPOM> and Android<https://github.com/notifications/mobile/android/AAYSHRTMD2HIMZ334SPNSKL4XPQRBA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNRXGA3TEMJRUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVZTG633UMVZF6YLOMRZG62LE>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[KristaMHill]

Just took a look in ack/.obj/modules/src/alloc/+alloc_hdr and found the dir to be empty, so perhaps rm did delete the first item listed in the error, then choked on what follows.

[KristaMHill]

Just for the record, just tried ninja again in the command envisonment
C:\Users\Hill>ninja
'ninja' is not recognized as an internal or external command,
operable program or batch file.

[KristaMHill]

I'm very much a newbie... Just found a reference to pass along:

[[ Ninja errors involving && typically occur because ninja executes commands directly rather than through a shell, meaning shell operators like && (and ||, >, |) are not interpreted correctly in raw command strings. To fix this, explicitly invoke a shell to parse the offending command string. ]]

Just looked in .obj/build.ninja and see that many lines that look like the commands in the errors we saw, that don't invoke a shell.

[KristaMHill]

I found an update on how to prepare for building the ACK under Win MSYS2. I'm using UCRT64. While this note may help others wanting to build the ACK, my result is as I've seen before, in this thread. The gh command requires credentials and didn't help the build, but is from a prior note in this thread

'Best,
Krista

===========
pacman -Syu
pacman -S --needed base-devel mingw-w64-x86_64-toolchain flex bison python3 lua
git clone https://github.com/davidgiven/ack.git
cd ack
gh pr checkout 364
make

[KristaMHill]

just to add, I'm not sure if ninja is included in base-devel, but ninja is needed