Re: bin/59224: Build failure on Solaris-11.4 for sh/init.c

[Sad Clouds <cryintothebluesky%gmail.com@localhost>]

On Wed, 26 Mar 2025 23:10:02 +0000 (UTC)
"Robert Elz via gnats" <gnats-admin%NetBSD.org@localhost> wrote:

>  That's weird, though Solaris has been known to have some broken shells at
>  times.

That was my first suspicion, although debugging it is proving a bit
difficult:

mkdir /tmp/test && cd /tmp/test

AWK=/opt/netbsd/tools.sparc64/bin/nbawk  MKTEMP=/opt/netbsd/tools.sparc64/bin/nbmktemp  SED=/opt/netbsd/tools.sparc64/bin/nbsed /bin/sh /opt/netbsd/netbsd-10-src/bin/sh/mkinit.sh /opt/netbsd/netbsd-10-src/bin/sh/alias.c /opt/netbsd/netbsd-10-src/bin/sh/arith_token.c /opt/netbsd/netbsd-10-src/bin/sh/arithmetic.c /opt/netbsd/netbsd-10-src/bin/sh/cd.c /opt/netbsd/netbsd-10-src/bin/sh/bltin/echo.c /opt/netbsd/netbsd-10-src/bin/sh/error.c /opt/netbsd/netbsd-10-src/bin/sh/eval.c /opt/netbsd/netbsd-10-src/bin/sh/exec.c /opt/netbsd/netbsd-10-src/bin/sh/expand.c /opt/netbsd/netbsd-10-src/bin/sh/histedit.c /opt/netbsd/netbsd-10-src/bin/sh/input.c /opt/netbsd/netbsd-10-src/bin/sh/jobs.c /opt/netbsd/netbsd-10-src/bin/sh/mail.c /opt/netbsd/netbsd-10-src/bin/sh/main.c /opt/netbsd/netbsd-10-src/bin/sh/memalloc.c /opt/netbsd/netbsd-10-src/bin/sh/miscbltin.c /opt/netbsd/netbsd-10-src/bin/sh/mystring.c /opt/netbsd/netbsd-10-src/bin/sh/options.c /opt/netbsd/netbsd-10-src/bin/sh/parser.c /opt/netbsd/net
 bsd-10-src/bin/sh/redir.c /opt/netbsd/netbsd-10-src/bin/sh/show.c /opt/netbsd/netbsd-10-src/bin/sh/trap.c /opt/netbsd/netbsd-10-src/bin/sh/output.c /opt/netbsd/netbsd-10-src/bin/sh/var.c /opt/netbsd/netbsd-10-src/bin/test/test.c /opt/netbsd/netbsd-10-src/bin/kill/kill.c /opt/netbsd/netbsd-10-src/bin/sh/syntax.c

and then looking at the generate init.c, no issues there with the generated code.

        /* from /opt/netbsd/netbsd-10-src/bin/sh/trap.c: */
        {
                char *sm;

                INTOFF;
                clear_traps(2);
                for (sm = sigmode ; sm < sigmode + NSIG ; sm++) {
                        if (*sm == S_IGN) {
                                *sm = S_HARD_IGN;
                                VTRACE(DBG_TRAP, ("SHELLPROC: %d -> hard_ign\n",
                                    (sm - sigmode)));
                        }
                }
                INTON;
        }

I will run the build a few more times and see if the issue appears
again. If not, then I assume this was a weird /bin/sh runtime bug.