Implement support for POSIX_SPAWN_TCSETPGROUP and fix spawnveg bugs

[JohnoKing]

This commit implements support for the glibc POSIX_SPAWN_TCSETPGROUP and posix_spawnattr_tcsetpgrp_np extensions¹. This was done with the intention of improving performance in interactive shells without reintroducing previous race conditions²³. These changes were tested using glibc commit 342cc934.

src/cmd/ksh93/sh/path.c:
- Tell spawnveg to set the terminal process group when launching a child process in an interactive shell.

src/cmd/ksh93/sh/xec.c:
- Allow usage of spawnveg when posix_spawnattr_tcsetpgrp_np is available.
- Reimplement most of the sh_ntfork() SIGTSTP handling code removed in commit 66c3720.

src/lib/libast/comp/spawnveg.c,
src/lib/libast/misc/procopen.c,
src/lib/libast/features/sys:
- Add support for POSIX_SPAWN_TCSETPGROUP as implemented in glibc. This code (with a few changes) might also work on QNX, but I've avoided changing anything on that OS since I don't have a QNX system to test with.
- Bugfix 1: If pgid is -1, use POSIX_SPAWN_SETSID when it's available (spawnveg uses setsid(2) in the fork fallback, so it should do the same thing when posix_spawn is used instead).

src/lib/libast/man/spawnveg.3:
- Document the changes to spawnveg in the corresponding man page. Currently spawnveg will only use the new tcfd argument if posix_spawnattr_tcsetpgrp_np is supported. This could also be implemented for the fork fallback, but for now I've avoided changing that since actually using it in the fork code would likely require a lot of hackery to avoid attempting tcsetpgrp with vfork (the behavior of tcsetpgrp after vfork is not portable) and would only benefit systems that don't have posix_spawn and vfork (I can't recall any off the top of my head that would fall under that category).

src/lib/libast/features/lib:
- Bugfix 2: Fix syntax errors in the posix_spawn feature test that caused ksh to fall back to vfork instead.
- Detect the existence of the GNU posix_spawnattr_tcsetpgrp_np() extension.

[McDutchie]

How would I go about testing this in a way that is not massively time-consuming? Replacing glibc on an existing distro generally breaks things, doesn't it?

Since these changes are new to both glibc and ksh, it's probably best to keep them to the dev branch only, so that they can see a good amount of testing before making it into 1.1, when that time comes.

I'd like to separate the bugfixes into a commit of their own that I can also backport to the 1.0 branch – particularly the syntax errors in features/lib which are my fault, I introduced them in 1064933 – oops. That was a sed gone wrong. I'm going to have to go over that commit again with a fine-toothed comb to check if any other problems were introduced.

[JohnoKing]

I'm not entirely sure of how to go about replacing glibc on most distributions. I'm currently using the glibc-git AUR package to test these changes. The PKGBUILD script could be used as a reference for building and installing glibc using the latest git commit.

I'd like to separate the bugfixes into a commit of their own that I can also backport to the 1.0 branch

The other bugfix in this pull request is the change to use POSIX_SPAWN_SETSID when it's available (so spawnveg when using posix_spawn acts the same as when using the fork fallback). I'll separate that into it's own pull request (edit: now #439).

[McDutchie]

So, I looked up how AUR packages work and finally got as far as running makepkg on that glibc-git AUR package, and now the build fails:

I really don't have time to mess with this any more, I'll just take your word for it all. If it breaks, well, it's the dev branch.

[JohnoKing]

I figured out that the build failure you encountered is caused by the default CFLAGS in /etc/makepkg.conf (I hadn't encountered that failure on my system because my makepkg.conf had different CFLAGS). This should fix the glibc-git build failure:

--- makepkg.conf	2022-01-08 01:13:06.000000000 -0800
+++ makepkg.conf.patch	2022-02-03 11:16:59.612926347 -0800
@@ -38,9 +38,10 @@
 
 #-- Compiler and Linker Flags
 #CPPFLAGS=""
-CFLAGS="-march=x86-64 -mtune=generic -O2 -pipe -fno-plt -fexceptions \
-        -Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \
-        -fstack-clash-protection -fcf-protection"
+#CFLAGS="-march=x86-64 -mtune=generic -O2 -pipe -fno-plt -fexceptions \
+#        -Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \
+#        -fstack-clash-protection -fcf-protection"
+CFLAGS="-march=x86-64 -O2 -pipe "
 CXXFLAGS="$CFLAGS -Wp,-D_GLIBCXX_ASSERTIONS"
 LDFLAGS="-Wl,-O1,--sort-common,--as-needed,-z,relro,-z,now"
 #RUSTFLAGS="-C opt-level=2"

Additionally, glibc 2.35 has recently been released (changelog), although right before the 2.35 release posix_spawnattr_tcsetpgrp_np was replaced by posix_spawn_file_actions_addtcsetpgrp_np. I'm currently testing a patch that updates spawnveg in accordance with that change.