Re: pkg/59002: Bootstrapping pkgsrc on HP-UX 11.11/PA-RISC fails due to lack of stdint.h

["David H. Gutteridge via gnats" <gnats-admin%NetBSD.org@localhost>]

The following reply was made to PR pkg/59002; it has been noted by GNATS.

From: "David H. Gutteridge" <david%gutteridge.ca@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: pkg/59002: Bootstrapping pkgsrc on HP-UX 11.11/PA-RISC fails due
 to lack of stdint.h
Date: Thu, 06 Feb 2025 00:21:32 -0500

 Hi,
 
 There are two different ways to look at this, as I see it. You
 mentioned you aren't able to get a full error log. Please try using
 script(1) to capture the full output to a log file. It should be
 available on your HP-UX release (I recall using it on 11.-something-
 or-other a long time ago). Just "script [logfile name]" then restart
 the same build process as before, then "exit" the next time there's a
 prompt, and there's your log file. (Which if huge in size would be
 better posted somewhere other than NetBSD Gnats, perhaps.)
 
 Having said that, I'm not sure we need to capture *everything* to
 examine what you shared so far. On the surface, this seems a slightly
 odd error to me. It doesn't help that the output from bmake doesn't
 include the error code, and the mapping of string to code doesn't seem
 to be publicly available. If it's not too much trouble, could you
 please do the following?
 
 1. Apply the patch below and recompile bmake, so we will get the
 actual error code occurring (it can't be the normal "no match found",
 represented by REG_NOMATCH, as that's already accounted for by
 necessity).
 2. Re-run the failing build step (probably using script, too).
 3. Go to /usr/include/regex.h (or wherever it's found in your
 installation), and find that actual value among the REG_* error codes
 that (should be) found there. That should give us more of a clue what
 it finds objectionable. You'd start looking for the numeric value I've
 added in brackets to the error: find REG_NOMATCH and work down from
 there in the header file to map the code backwards.
 
 Index: var.c
 ===================================================================
 RCS file: /cvsroot/pkgsrc/devel/bmake/files/var.c,v
 retrieving revision 1.14
 diff -u -r1.14 var.c
 --- var.c	17 Sep 2024 11:52:27 -0000	1.14
 +++ var.c	6 Feb 2025 05:05:55 -0000
 @@ -1578,7 +1578,7 @@
   	size_t errlen = regerror(reerr, pat, NULL, 0);
   	char *errbuf = bmake_malloc(errlen);
   	regerror(reerr, pat, errbuf, errlen);
 -	Parse_Error(PARSE_FATAL, "%s: %s", str, errbuf);
 +	Parse_Error(PARSE_FATAL, "%s: %s (%d)", str, errbuf, reerr);
   	free(errbuf);
   }
 
 (As I'm sure you know, it's kind of hard to support this stuff without
 access to the environment. I used to work on HP-UX years ago, but it's
 been a while.) Bootstrapping pkgsrc on HP-UX PA-RISC was reported to
 work at one point. I don't see anything in the problem make code that
 has changed appreciably over a long period of time.
 
 I do have PA-RISC hardware myself (a B180L+), but it is running NetBSD.
 I could try to follow part of your journey in parallel, but would only
 hit a percentage of the same issues. (I do remember submitting patches
 for individual packages breaking under PA-RISC some years ago, but
 don't recall what they were now. Some projects have probably moved on
 in coding assumptions since then.)
 
 Regards,
 
 Dave