macOS: Some environment variables are not working for alternatives

[bartoszkosiorek]

Due to System Integrity Protection (SIP) on macOS systems, the binaries from the /bin (eg. /bin/bash) are not allowed to handle this environment variables. More information: https://support.apple.com/en-us/HT204899

Example of SIP:

• Create ./test.sh file

#!/bin/bash
echo $DYLD_LIBRARY_PATH

• Run with command: DYLD_LIBRARY_PATH=/lalala ./test.sh

Result:

• Nothing is printed:
  $ DYLD_LIBRARY_PATH=/lalala ./test.sh

The current generated alternative script files on macOS from the pkg_alternatives package are having as interpreter /bin/bash, eg. pip3:

$ head /usr/local/pkg/bin/pip3
#!/bin/bash
#
# $NetBSD: wrapper.sh,v 1.2 2012/06/13 15:35:32 jperkin Exp $
#
# pkg_alternatives - Generic wrappers for programs with similar interfaces
# Copyright (c) 2005 Julio M. Merino Vidal <jmmv@NetBSD.org>
...

This is a bit problematic in case of macOS as due to SIP protection for example declared environment variable like DYLD_LIBRARY_PATH is rejected.

This will lead eventually to problems for trying to do things like DYLD_LIBRARY_PATH=/path_with_dylibs /usr/local/pkg//bin/ctest which is pretty common.

[jperkin]

I think I've seen this worked around by using a shell from pkgsrc, we may want to investigate going back to bootstrapping our own and using it for everything.

[bartoszkosiorek]

For my own project I would like to change the default shebang for wrappers.
It seems that it could be modified in:
https://github.com/NetBSD/pkgsrc/blob/trunk/pkgtools/pkg_alternatives/files/wrapper.sh#L1

Do you know by which option the __SH__ or @SH@ could be set?

[bsiegert]

I think it is set to the value of the SHELL environment variable during bootstrap? Bartosz <notifications@github.com> schrieb am Fr., 12. Juni 2020, 16:13:

…

For my own project I would like to change the default shebang for wrappers. It seems that it could be modified in: https://github.com/NetBSD/pkgsrc/blob/trunk/pkgtools/pkg_alternatives/files/wrapper.sh#L1 Do you know by which option the __SH__ or @sh@ could be set? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#66 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGTQGVY6SWRFL4KFGVY6R3RWIZYVANCNFSM4N4F3IVA> .

[jperkin]

Simplest way is:

$ cd pkgsrc/pkgtools/pkg_alternatives
$ bmake TOOLS_PLATFORM.sh=/opt/pkg/bin/bash install

or whatever shell you want to use outside of /bin.

[bartoszkosiorek]

@jperkin I have found also WRAPPER_SH variable in settings:
https://github.com/NetBSD/pkgsrc/blob/pkgsrc-2019Q1/mk/wrapper/wrapper-defs.mk#L41

[jperkin]

No, that's something completely different (and something you're unlikely to be using).

[bartoszkosiorek]

What do you think about idea, of use bash from pkgsrc/shells/bash package by pkg_alternatives for wrappers?

It will resolve such kind of issues, but it will be additional dependency to shells/bash.

[jperkin]

As a local change it would be fine, nothing depends on pkg_alternatives so you aren't running any risks during upgrades etc.

For a wider pkgsrc fix we'll need to look into what the current status of shells/pdksh is on Darwin by running some bulk builds, and then if it looks ok consider switching bootstrap over to using it by default.

[bartoszkosiorek]

After set TOOLS_PLATFORM.sh=/usr/local/pkg/bin/bash, the wrappers have changed shebang correctly.

I will try to overwrite TOOLS_PLATFORM.sh variable in pkg_alternatives/Makefile and add dependency, to use the default shells/bash shell.

I have build the shells/pdksh without problems.

Logs from shells/pdksh build on macOS Catalina 10.15.5

Click to expand!

pdksh build logs

$ sudo /usr/local/pkg/bin/bmake deinstall
===> Deinstalling for pdksh-5.2.14nb7
Running /usr/local/pkg/sbin/pkg_delete -K /usr/local/pkg/pkgdb  pdksh-5.2.14nb7
$ sudo /usr/local/pkg/bin/bmake clean
===> Cleaning for pdksh-5.2.14nb7
$ sudo /usr/local/pkg/bin/bmake install
===> Installing dependencies for pdksh-5.2.14nb7
=> Build dependency cwrappers>=20150314: found cwrappers-20180325
===> Overriding tools for pdksh-5.2.14nb7
===> Extracting for pdksh-5.2.14nb7
/bin/cp -R /Users/NavKit/dev/pkgsrc/shells/pdksh/files /private/var/tmp/pkgsrc-obj/shells/pdksh/work/pdksh-5.2.14
===> Patching for pdksh-5.2.14nb7
===> Creating toolchain wrappers for pdksh-5.2.14nb7
===> Configuring for pdksh-5.2.14nb7
=> Modifying GNU configure scripts to avoid --recheck
=> Replacing config-guess with pkgsrc versions
=> Replacing config-sub with pkgsrc versions
=> Replacing install-sh with pkgsrc version
creating cache ./config.cache
checking for gcc... clang
checking whether we are using GNU C... yes
checking how to run the C preprocessor... clang -E
checking whether clang needs -traditional... no
checking if this is a problematic os... checking for minix/config.h... no
no
checking for dirent.h that defines DIR... yes
checking for opendir in -ldir... no
checking for opendir in -lndir... no
checking for sane unistd.h... yes
checking terminal interface... termios
checking for stddef.h... yes
checking for stdlib.h... yes
checking for string.h... yes
checking for memory.h... yes
checking for fcntl.h... yes
checking for limits.h... yes
checking for paths.h... yes
checking for sys/param.h... yes
checking for sys/resource.h... yes
checking for values.h... no
checking for ulimit.h... yes
checking for sys/time.h... yes
checking whether time.h and sys/time.h may both be included... yes
checking for sys/wait.h that is POSIX.1 compatible... yes
checking for off_t in sys/types.h... yes
checking for mode_t in sys/types.h... yes
checking for pid_t in sys/types.h... yes
checking for uid_t in sys/types.h... yes
checking return type of signal handlers... void
checking size of int... 4
checking size of long... 8
checking for clock_t in any of <sys/types.h>, <sys/times.h> and <sys/time.h>... yes
checking for sigset_t in <sys/types.h> and <signal.h>... yes
checking for rlim_t in <sys/types.h> and <sys/resource.h>... yes
checking for working memmove... yes
checking for memset... yes
checking for confstr... yes
checking for dup2... yes
checking for flock... yes
checking for getcwd... yes
checking for getwd... yes
checking for killpg... yes
checking for nice... yes
checking for setrlimit... yes
checking for strerror... yes
checking for strcasecmp... yes
checking for strstr... yes
checking for sysconf... yes
checking for tcsetpgrp... yes
checking for ulimit... yes
checking for waitpid... yes
checking for wait3... yes
checking for strlcpy... yes
checking for strlcat... yes
checking for sigsetjmp... yes
checking for valloc... yes
checking for getpagesize... yes
checking for working mmap... yes
checking for lstat... yes
checking for sys_errlist declaration in errno.h... no
checking for sys_errlist in library... yes
checking for sys_siglist declaration in signal.h or unistd.h... yes
checking for sys_siglist in library... yes
checking time() declaration in time.h... yes
checking if times() is present/working... yes
checking whether stat file-mode macros are broken... no
checking for st_rdev in struct stat... yes
checking for working const... yes
checking if compiler understands void... yes
checking if compiler understands volatile... yes
checking if compiler understands prototypes... yes
checking if C compiler groks __attribute__(( .. ))... yes
checking whether #! works in shell scripts... yes
checking for a BSD compatible install... /usr/bin/install -c -o root -g wheel
checking if dup2() works (ie, resets the close-on-exec flag)... yes
checking flavour of signal routines... posix
checking flavour of pgrp routines... posix
checking if process group synchronization is required... no
checking if opendir() fails to open non-directories... yes
checking if you have /dev/fd/n... yes
updating cache ./config.cache
creating ./config.status
creating Makefile
creating config.h
===> Building for pdksh-5.2.14nb7
--- emacs.out ---
--- ksh.1 ---
--- stamp-h ---
--- emacs.out ---
./emacs-gen.sh ./emacs.c > tmpemacs.out
--- ksh.1 ---
./mkman ksh ./ksh.Man > tmpksh.1
--- stamp-h ---
CONFIG_FILES="" CONFIG_HEADERS=config.h ./config.status
--- emacs.out ---
mv tmpemacs.out emacs.out
--- ksh.1 ---
mv tmpksh.1 ksh.1
--- stamp-h ---
creating config.h
config.h is unchanged
date > stamp-h
--- siglist.out ---
--- alloc.o ---
--- c_ksh.o ---
--- c_sh.o ---
--- c_test.o ---
--- c_ulimit.o ---
--- edit.o ---
--- emacs.o ---
--- siglist.out ---
./siglist.sh "clang -E -P  -DHAVE_CONFIG_H -I. -I." < ./siglist.in > tmpsiglist.out
--- alloc.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 alloc.c
--- c_ksh.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 c_ksh.c
--- c_sh.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 c_sh.c
--- c_test.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 c_test.c
--- c_ulimit.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 c_ulimit.c
--- edit.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 edit.c
--- emacs.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 emacs.c
--- c_ksh.o ---
c_ksh.c:571:13: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                        shprintf(newline);
                                 ^~~~~~~
c_ksh.c:571:13: note: treat the string as an argument to avoid this
                        shprintf(newline);
                                 ^
                                 "%s", 
c_ksh.c:871:17: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                            shprintf(newline);
                                     ^~~~~~~
c_ksh.c:871:17: note: treat the string as an argument to avoid this
                            shprintf(newline);
                                     ^
                                     "%s", 
c_ksh.c:962:14: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                shprintf(newline);
                                         ^~~~~~~
c_ksh.c:962:14: note: treat the string as an argument to avoid this
                                shprintf(newline);
                                         ^
                                         "%s", 
c_ksh.c:986:14: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                shprintf(newline);
                                         ^~~~~~~
c_ksh.c:986:14: note: treat the string as an argument to avoid this
                                shprintf(newline);
                                         ^
                                         "%s", 
c_ksh.c:1234:13: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                bi_errorf(null);
                          ^~~~
c_ksh.c:1234:13: note: treat the string as an argument to avoid this
                bi_errorf(null);
                          ^
                          "%s", 
c_ksh.c:1255:13: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                        shprintf(newline);
                                 ^~~~~~~
c_ksh.c:1255:13: note: treat the string as an argument to avoid this
                        shprintf(newline);
                                 ^
                                 "%s", 
--- eval.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 eval.c
--- exec.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 exec.c
--- siglist.out ---
mv tmpsiglist.out siglist.out
--- expr.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 expr.c
--- eval.o ---
eval.c:340:19: warning: implicit conversion from 'int' to 'char' changes value from 192 to -64 [-Wconstant-conversion]
                                                *dp++ = '@' + 0x80;
                                                      ~ ~~~~^~~~~~
--- history.o ---
--- exec.o ---
exec.c:148:13: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                        errorf(null);
                                               ^~~~
exec.c:148:13: note: treat the string as an argument to avoid this
                                        errorf(null);
                                               ^
                                               "%s", 
--- history.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 history.c
--- expr.o ---
expr.c:190:11: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                        errorf(null);
                               ^~~~
expr.c:190:11: note: treat the string as an argument to avoid this
                        errorf(null);
                               ^
                               "%s", 
--- io.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 io.c
--- expr.o ---
1 warning generated.
--- jobs.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 jobs.c
--- lex.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 lex.c
--- c_ksh.o ---
6 warnings generated.
--- mail.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 mail.c
--- jobs.o ---
jobs.c:897:11: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
        shprintf(newline);
                 ^~~~~~~
jobs.c:897:11: note: treat the string as an argument to avoid this
        shprintf(newline);
                 ^
                 "%s", 
--- main.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 main.c
--- misc.o ---
--- lex.o ---
lex.c:858:9: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
        errorf(null);
               ^~~~
lex.c:858:9: note: treat the string as an argument to avoid this
        errorf(null);
               ^
               "%s", 
--- misc.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 misc.c
--- main.o ---
main.c:562:13: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                        shellf(newline);
                                               ^~~~~~~
main.c:562:13: note: treat the string as an argument to avoid this
                                        shellf(newline);
                                               ^
                                               "%s", 
main.c:861:9: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
        errorf(null); /* this is never executed - keeps gcc quiet */
               ^~~~
main.c:861:9: note: treat the string as an argument to avoid this
        errorf(null); /* this is never executed - keeps gcc quiet */
               ^
               "%s", 
--- exec.o ---
1 warning generated.
--- missing.o ---
--- path.o ---
--- missing.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 missing.c
--- path.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 path.c
--- misc.o ---
misc.c:262:12: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                shprintf(newline);
                         ^~~~~~~
misc.c:262:12: note: treat the string as an argument to avoid this
                shprintf(newline);
                         ^
                         "%s", 
--- eval.o ---
1 warning generated.
--- misc.o ---
misc.c:1014:15: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                bi_errorf(null);
                                          ^~~~
misc.c:1014:15: note: treat the string as an argument to avoid this
                                bi_errorf(null);
                                          ^
                                          "%s", 
misc.c:1040:15: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                bi_errorf(null);
                                          ^~~~
misc.c:1040:15: note: treat the string as an argument to avoid this
                                bi_errorf(null);
                                          ^
                                          "%s", 
misc.c:1091:20: warning: adding 'int' to a string does not append to the string [-Wstring-plus-int]
                        shprintf("'\\'" + 1 - inquote);
                                 ~~~~~~~^~~
misc.c:1091:20: note: use array indexing to silence this warning
                        shprintf("'\\'" + 1 - inquote);
                                        ^
                                 &      [  ]
misc.c:1091:13: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                        shprintf("'\\'" + 1 - inquote);
                                 ^~~~~~~~~~~~~~~~~~~~
misc.c:1091:13: note: treat the string as an argument to avoid this
                        shprintf("'\\'" + 1 - inquote);
                                 ^
                                 "%s", 
--- shf.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 shf.c
--- sigact.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 sigact.c
--- main.o ---
2 warnings generated.
--- syn.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 syn.c
--- table.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 table.c
--- tree.o ---
--- jobs.o ---
1 warning generated.
--- tree.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 tree.c
--- tty.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 tty.c
--- var.o ---
--- version.o ---
--- var.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 var.c
--- version.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 version.c
--- vi.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 vi.c
--- misc.o ---
5 warnings generated.
--- trap.o ---
clang -c  -DHAVE_CONFIG_H -I. -I. -O2 trap.c
--- var.o ---
var.c:373:11: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                        errorf(null);
                               ^~~~
var.c:373:11: note: treat the string as an argument to avoid this
                        errorf(null);
                               ^
                               "%s", 
var.c:727:14: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                    errorf(null);
                           ^~~~
var.c:727:14: note: treat the string as an argument to avoid this
                    errorf(null);
                           ^
                           "%s", 
--- vi.o ---
vi.c:1761:6: warning: address of array 'newline' will always evaluate to 'true' [-Wpointer-bool-conversion]
        if (newline) {
        ~~  ^~~~~~~
--- lex.o ---
1 warning generated.
--- var.o ---
2 warnings generated.
--- vi.o ---
1 warning generated.
--- ksh ---
clang -L/usr/local/pkg/lib -o ksh alloc.o c_ksh.o c_sh.o c_test.o c_ulimit.o edit.o emacs.o  eval.o exec.o expr.o history.o io.o jobs.o lex.o mail.o  main.o misc.o missing.o path.o shf.o sigact.o syn.o table.o trap.o  tree.o tty.o var.o version.o vi.o 
===> Installing for pdksh-5.2.14nb7
=> Generating pre-install file lists
=> Creating installation directories
/usr/bin/install -c  -o root -g wheel -m 755 /private/var/tmp/pkgsrc-obj/shells/pdksh/work/pdksh-5.2.14/ksh /private/var/tmp/pkgsrc-obj/shells/pdksh/work/.destdir/usr/local/pkg/bin/pdksh
/usr/bin/install -c -o root -g wheel -m 644 /private/var/tmp/pkgsrc-obj/shells/pdksh/work/pdksh-5.2.14/ksh.1 /private/var/tmp/pkgsrc-obj/shells/pdksh/work/.destdir/usr/local/pkg/man/man1/pdksh.1
=> Automatic stripping binaries
=> Automatic manual page handling
=> Generating post-install file lists
=> Checking file-check results for pdksh-5.2.14nb7
=> Creating binary package /private/var/tmp/pkgsrc-obj/shells/pdksh/work/.packages/pdksh-5.2.14nb7.tgz
===> Building binary package for pdksh-5.2.14nb7
=> Creating binary package /Users/NavKit/dev/pkgsrc/packages/All/pdksh-5.2.14nb7.tgz
===> Installing binary package of pdksh-5.2.14nb7
===========================================================================
The following lines can be added to /etc/shells:

	/usr/local/pkg/bin/pdksh

===========================================================================
$  /usr/local/pkg/bin/pdksh
$ exit
$ uname -a
Darwin pl1mcl-5287921 19.5.0 Darwin Kernel Version 19.5.0: Tue May 26 20:41:44 PDT 2020; root:xnu-6153.121.2~2/RELEASE_X86_64 x86_64

[bartoszkosiorek]

@jperkin Do you know how to setup shell from pkgsrc globally in bootstrap?
The bootstrap tools are defined here:
https://github.com/NetBSD/pkgsrc/blob/trunk/mk/tools/tools.Darwin.mk#L100

How these paths needs to be set properly to use pkgsrc shell and how add dependency to boostrap?

[jperkin]

Yes, I'm currently testing this in a bulk build.

[bartoszkosiorek]

Could you please share the code with updated bootstrap?
I would like to start my own testing.

[jperkin]

For testing I'm just using bootstrap with the --full argument, this ensures that pdksh is bootstrapped and set as the default shell. If it looks ok then the permanent fix would be adding need_ksh=true to the Darwin setup section, probably being limited to OS releases that ship with SIP.

[bartoszkosiorek]

@jperkin Do you know if the zsh could be used by default for macOS/Darwin, as it is default shell:
https://www.theverge.com/2019/6/4/18651872/apple-macos-catalina-zsh-bash-shell-replacement-features
?

[jperkin]

One of the build hosts crashed over the weekend (Catalina is highly unreliable), I'll need to restart it.

As for zsh, I wouldn't recommend it, it's not designed to be a fast POSIX-only shell, but more as a feature-full interactive shell.

[bartoszkosiorek]

I think I found the root cause of my problems:
During run bootstrap, I was using SH=/bin/bash environment variable:
SH=/bin/bash ./bootstrap --prefix "$PKGSRC_PREFIX" --prefer-pkgsrc yes --workdir "$WORK_DIR" $__UNPRIVILEGED

After removing SH=/bin/bash, and use default shell, most/all of alternatives are using #!/usr/local/pkg/bin/bash shebang (/usr/local/pkg/ is a directory where pkgsrc is installed).

Do you know why it choose the bash instead of sh, as defined here?
https://github.com/NetBSD/pkgsrc/blob/trunk/mk/tools/tools.Darwin.mk#L100

The /usr/local/pkg/etc/mk.conf content:

# Example /usr/local/pkg/etc/mk.conf file produced by bootstrap-pkgsrc
# Tue Jun 16 11:12:26 CEST 2020
.ifdef BSD_PKG_MK       # begin pkgsrc settings
ABI=                    64
PKGSRC_COMPILER=        clang
CC=                     clang
CXX=                    clang++
CPP=                    ${CC} -E
CLANGBASE=              /usr
UNPRIVILEGED=           yes
PKG_DBDIR=              /usr/local/pkg/pkgdb
LOCALBASE=              /usr/local/pkg
VARBASE=                /usr/local/pkg/var
PKG_TOOLS_BIN=          /usr/local/pkg/sbin
PKGINFODIR=             info
PKGMANDIR=              man
PREFER_PKGSRC=          yes
MAKE_JOBS=8
SKIP_LICENSE_CHECK=yes
WRKOBJDIR=/private/var/tmp/pkgsrc-obj
X11_TYPE=modular
RUBY_VERSION_DEFAULT=25
ALLOW_VULNERABLE_PACKAGES=yes
PREFER_PKGSRC=  yes
PKG_OPTIONS.libfetch+= openssl
INSTALL_UNSTRIPPED=no
STRIP_BINARIES=yes
PKGSRC_USE_RELRO=full
BUILDLINK_USE_HARDLINKS=no
.endif                  # end pkgsrc settings

[bartoszkosiorek]

@jperkin Do you find any issue with building ksh by default in bootstrap?

[jperkin]

Bootstrap will default to /bin/sh (i.e. bash) unless you override SH in the bootstrap environment. If it's using /usr/local/pkg/bin/bash then that will be from a setting you have somewhere (you mentioned earlier about patching things so it's probably from that).

The bulk build isn't finished yet, they usually take around 3 days to complete on macOS from scratch.

[jperkin]

FWIW it's currently at 17380/22794, producing 13451 packages, so looks reasonable so far.

[bartoszkosiorek]

Thanks. I found an issue in my internal scripts.

After enabling need_ksh on Darwin, the shell used by pkgsrc (pdksh) will be completely independent from what is installed on system (etc. /bin/bash or /bin/zsh).

So it will resolve issue with:
https://support.apple.com/en-us/HT208050

Apple could remove /bin/bash and /bin/zsh and the pkgsrc will be still working with shells/pdksh.

Am I right?

[jperkin]

I've posted the results of the bulk build along with a patch for bootstrap here for review: http://mail-index.netbsd.org/tech-pkg/2020/06/18/msg023387.html

[bartoszkosiorek]

@jperkin Thanks for your patch!

Could you please enable need_ksh=yes as default, and disable it only on old systems?
Otherwise the versions needs to be updated once the new macOS version will be released.

Something similar to:
https://github.com/NetBSD/pkgsrc/pull/67/files

[bartoszkosiorek]

@jperkin I think every package, which use pkg_alternatives, should have running dependency on shells/pdksh shell. Otherwise, it could happen that the application wrapper couldn't be run due to missing pdksh.

[jperkin]

That would already be handled on account of it being a BOOTSTRAP_PKG.

[bartoszkosiorek]

After setting need_ksh, almost all +INSTALL scripts are using default shell (for macOS it will be pdksh).
For example pkg_alternatives:
https://github.com/NetBSD/pkgsrc/blob/trunk/pkgtools/pkg_alternatives/INSTALL

But for some packages, the shell is hardcoded. For example for pkg_install:
https://github.com/NetBSD/pkgsrc/blob/trunk/pkgtools/pkg_install/INSTALL#L1

Do you know the reason for that? Why the default shell is not used for pkg_install?

[jperkin]

I think just an oversight, though it's probably a good idea to be as safe as possible for critical packages and there's nothing in the pkg_install INSTALL script that would cause problems for any Bourne shell.

[bartoszkosiorek]

What else needs to be done to enable ksh by default for macOS bootstrap?

[jperkin]

After running into a separate issue in pdksh I went an alternative route and implemented mksh as a bootstrap shell. I'm just waiting on review of http://mail-index.netbsd.org/tech-pkg/2020/06/29/msg023478.html and will, once that's imported, switch macOS over to it.