Re: head -r333946(?) broke many builds

On 2018-May-20, at 7:35 PM, Mark Millard <marklmi26-fbsd at yahoo.com> wrote:

> ci.freebsd.org shows -r333945 built okay (ignoring riscv64) but
> later not ( -r333947 and later for most).
> 
> --- all_subdir_usr.bin/top ---
> make[4]: make[4]: don't know how to make top.local.h. Stop
> 
> make[4]: stopped in /usr/src/usr.bin/top
> *** [all_subdir_usr.bin/top] Error code 2
> 
> 
>> Author: eadler
>> Date: Sun May 20 23:37:30 2018
>> New Revision: 333946
>> URL: 
>> https://svnweb.freebsd.org/changeset/base/333946
>> 
>> 
>> Log:
>>  top(1): pull configuration directly into header files
>> 
>> . . .
> 
> 
> 
> Interestingly FreeBSD-head-aarch64-build is not broken
> at -r333947 but is later at -r333949 --for the same
> error:
> 
> --- all_subdir_usr.bin ---
> make[4]: make[4]: don't know how to make top.local.h. Stop
> 
> make[4]: stopped in /usr/src/usr.bin/top
> *** [all_subdir_usr.bin/top] Error code 2
> 

In my own builds it looks like WITH_META_MODE= does not deal with
top having moved. The builds did not fail because they found and
used an old top.local.h in my /usr/obj/ areas for the various build
targets that I tried. For example:

# ls -lTt /usr/obj/amd64_clang/amd64.amd64/usr/src/amd64.amd64/usr.bin/top/
total 928
. . .
-rw-r--r--  1 root  wheel    1017 Nov  3 00:11:28 2017 top.local.h.meta
-rw-r--r--  1 root  wheel    2418 Nov  3 00:11:28 2017 top.local.h
FBSDFSSD# more /usr/obj/amd64_clang/amd64.amd64/usr/src/amd64.amd64/usr.bin/top/top.local.h.meta
# Meta data file /usr/obj/amd64_clang/amd64.amd64/usr/src/amd64.amd64/usr.bin/top/top.local.h.meta
CMD _at_echo Making top.local.h from /usr/src/contrib/top/top.local.hs
CMD _at_sed -e's,%LoadMax%,5.0,g'  -e's,%TableSize%,20011,g'  -e's,%NominalTopn%,18,g'  -e's,%topn%,-1,g'  -e's,%delay%,2,g'  -e's,%random%,1,g'  /usr/src/contrib/top/top.local.hs > top.local.h
CWD /usr/obj/amd64_clang/amd64.amd64/usr/src/amd64.amd64/usr.bin/top
TARGET top.local.h
-- command output --
Making top.local.h from /usr/src/contrib/top/top.local.hs

-- filemon acquired metadata --
# filemon version 5
# Target pid 10130
# Start 1509693088.899769
V 5
E 10133 /bin/sh
R 10133 /etc/libmap.conf
R 10133 /var/run/ld-elf.so.hints
R 10133 /lib/libedit.so.7
R 10133 /lib/libc.so.7
R 10133 /lib/libncursesw.so.8
F 10133 10134
W 10134 top.local.h
E 10134 /usr/bin/sed
R 10134 /etc/libmap.conf
R 10134 /var/run/ld-elf.so.hints
R 10134 /lib/libc.so.7
R 10134 /usr/src/contrib/top/top.local.hs
X 10134 0 0
X 10133 0 0
# Stop 1509693088.899769
# Bye bye

was still around for the #include's to find.

May be something like this explains FreeBSD-head-aarch64-build
initially not failing.

After a move (such as happened for top) forcing a full rebuild
"form scratch" seems to be required for testing if the build
process is working.

===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)