Date: Wed, 31 Oct 2001 21:49:25 -0800 (PST) From: David Wolfskill <david@catwhisker.org> To: current@FreeBSD.ORG, obrien@FreeBSD.ORG Subject: Re: buildworld breakage during "make depend" at usr.bin/kdump Message-ID: <200111010549.fA15nPG47227@bunrab.catwhisker.org> In-Reply-To: <200110312159.f9VLx1I45943@bunrab.catwhisker.org>
next in thread | previous in thread | raw e-mail | index | archive | help
OK; I was finally able to complete the buildworld/buildkernel/
installkernel/installworld/mergemaster cycle. In the process,
I found 3 awk scripts that required modification; I will append
the changes I made after another paragraph of my deathless prose. :-}
Before I get to the patches, I will temper the good news of being
able to build & install both the world and the kernel by admitting
that I was unable to actually boot the resulting system: The machine
(laptop; I have no monitor on the "build" machine) would prompt for
the boot slice, and whether I pressed the default selection (for the
newly-built world), pressed Enter, or let the selection default by
timeout, the screen would go blank, then show the BIOS splash screen
and re-prompt for a boot slice selection. The novelty wore off after
a few iterations, and so I switched to booting -STABLE. :-}
Here are the patches; I'll comment about them below:
Index: usr.bin/kdump/mkioctls
===================================================================
RCS file: /cvs/freebsd/src/usr.bin/kdump/mkioctls,v
retrieving revision 1.23
diff -u -r1.23 mkioctls
--- usr.bin/kdump/mkioctls 24 May 2001 08:47:57 -0000 1.23
+++ usr.bin/kdump/mkioctls 31 Oct 2001 21:27:35 -0000
@@ -25,10 +25,10 @@
egrep -v '^(netns)/' |
xargs egrep -l \
'^#[ ]*define[ ]+[A-Za-z_][A-Za-z0-9_]*[ ]+_IO[^a-z0-9_]' |
- sed -e 's/^/#include </' -e s'/$/>/'
+ awk '{printf("#include <%s>\\\\n", $1)}'
`
-echo "$ioctl_includes" |
+awk -v x="$ioctl_includes" 'BEGIN {print x}' |
gcc -E -I$1 -dM - |
awk -v ioctl_includes="$ioctl_includes" -v use_switch="$use_switch" '
BEGIN {
Index: gnu/lib/libregex/doc/Makefile
===================================================================
RCS file: /cvs/freebsd/src/gnu/lib/libregex/doc/Makefile,v
retrieving revision 1.4
diff -u -r1.4 Makefile
--- gnu/lib/libregex/doc/Makefile 11 Jan 1997 03:19:32 -0000 1.4
+++ gnu/lib/libregex/doc/Makefile 31 Oct 2001 23:39:13 -0000
@@ -5,7 +5,7 @@
CLEANFILES += regex.texi
regex.texi: xregex.texi ../regex.h
- awk -f ${.CURDIR}/include.awk -vsource=${.CURDIR}/../regex.h \
+ awk -f ${.CURDIR}/include.awk -v source=${.CURDIR}/../regex.h \
< ${.CURDIR}/xregex.texi \
| expand >$@
Index: sys/boot/common/merge_help.awk
===================================================================
RCS file: /cvs/freebsd/src/sys/boot/common/merge_help.awk,v
retrieving revision 1.2
diff -u -r1.2 merge_help.awk
--- sys/boot/common/merge_help.awk 28 Aug 1999 00:39:50 -0000 1.2
+++ sys/boot/common/merge_help.awk 1 Nov 2001 03:50:24 -0000
@@ -8,7 +8,7 @@
BEGIN \
{
state = 0;
- first = 0;
+ first = -1;
ind = 0;
}
@@ -20,13 +20,13 @@
}
# entry header
-/^# T[[:graph:]]+ (S[[:graph:]]+ )*D[[:graph:]][[:print:]]*$/ && (state == 1) \
+/^# T[\x21-\x7e]+ (S[\x21-\x7e]+ )*D[\x21-\x7e][\x20-\x7e]*$/ && (state == 1) \
{
- match($0, " T[[:graph:]]+");
+ match($0, " T[\x21-\x7e]+");
T = substr($0, RSTART + 2, RLENGTH - 2);
- match($0, " S[[:graph:]]+");
+ match($0, " S[\x21-\x7e]+");
S = (RLENGTH == -1) ? "" : substr($0, RSTART + 2, RLENGTH - 2);
- match($0, " D[[:graph:]][[:print:]]*$");
+ match($0, " D[\x21-\x7e][\x20-\x7e]*$");
D = substr($0, RSTART + 2);
# find a suitable place to store this one...
David O'Brien has suggested another (better, I think) approach for the
mkioctls patch. I'm reporting what I (had already) used.
For the regex/doc/Makefile, it seems that gawk allows the value for -v
to be adjacent to the -v itself, while the -CURRENT awk works better if
there is whitespace separating them. I cannot quote documentation to
support this, but appeal to empirical evidence.
Of the three, I think the merge_help.awk patch is the most troublesome.
First, the mode of failure appeared to be an endless output loop, each
cycle of which was very short. I've previously confessed my relative
lack of experience with awk, so I hope that folks will be charitable.
But in debugging the script, I found two things that (IMO) warranted
attention:
* It seems that while gawk supports the POSIX "bracket expressions"
(POSIXese for "character classes", basically), the awk we are now
using in -CURRENT does not appear to do so. (Again, I cite empirical
evidence, vs. documentation.)
As a result of this, none of the input lines were recognized in the
"state 1" expression on line 23. Thus, the "help" array was empty
when the script reached the end of the input.
* Since no lines matched, the "first" variable remained set at 0, and
since the help array was empty, the loop termination condition in the
END block could never be reached. I believe this to be suboptimal.
I found that by changing the intial value of "first" to -1, that made
the default condition for END loop *met* be default. Thus, with this
change, the mode of failure became slightly less obnoxious. (Ideally,
it would, in this case, output no lines at all, and possibly end with a
non-zero status code. I chose to defer that for a subsequent
exercise, should it be perceived to be desireable.)
Unfortunately, with the mode of failure of the boot process, I don't
really get much chance to figure out what is going on. I'll try again
in the morning.
Cheers,
david
--
David H. Wolfskill david@catwhisker.org
As a computing professional, I believe it would be unethical for me to
advise, recommend, or support the use (save possibly for personal
amusement) of any product that is or depends on any Microsoft product.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200111010549.fA15nPG47227>