Re: git: 557cc0f350e7 - main - bsd.man.mk: Handle MANSRC.{TARGET} for MK_MANCOMPRESS=no and empty MANBUILDCAT

On 8/28/25 08:10, Matteo Riondato wrote:
> 
> 
>> On Aug 27, 2025, at 4:12 PM, John Baldwin <jhb@freebsd.org> wrote:
>>
>> The branch main has been updated by jhb:
>>
>> URL: https://cgit.FreeBSD.org/src/commit/?id=557cc0f350e758a223dffe8bca359de9fc2642b2
>>
>> commit 557cc0f350e758a223dffe8bca359de9fc2642b2
>> Author:     John Baldwin <jhb@FreeBSD.org>
>> AuthorDate: 2025-08-27 20:11:54 +0000
>> Commit:     John Baldwin <jhb@FreeBSD.org>
>> CommitDate: 2025-08-27 20:11:54 +0000
>>
>>     bsd.man.mk: Handle MANSRC.{TARGET} for MK_MANCOMPRESS=no and empty MANBUILDCAT
>>
>>     Reviewed by:    kevans
>>     Fixes:          65f60d715fd9 ("bsd.man.mk: Add a MANSRC.{TARGET} variable")
>>     Sponsored by:   Chelsio Communications
>>     Differential Revision:  https://reviews.freebsd.org/D52181
>> ---
> 
> This did not fix the breakage of buildworld I reported. I had to disable “WITHOUT_MANCOMPRESS=y” from src.conf to make buildworld complete.

Locally for me it fixed libutil++ (which is what you reported), and while I also
tried a library that did not use MANSRC (libutil), something else must be going
on.

> Once I’ve done a successful buildworld with “WITHOUT_MANCOMPRESS=y” disabled, I tried to buildworld with “WITHOUT_MANCOMPRESS=y” enabled, and it failed again, at:
> 
> make[3]: don't know how to make apmd.8. Stop
> 
> (detailed message below)
> 
> I have “WITHOUT_APM=y” in src.conf, so I don’t know why it is trying to build ampd.8.

Hmm, that is because of this in usr.sbin/Makefile which I think is quite wrong
(it probably installs the manpage twice on i386 (and possibly in different packages?)
which would break pkgbase if we were building packages for i386):

# Add architecture-specific manpages
# to be included anyway
MAN=   apmd/apmd.8 \
        nvram/nvram.8

> I tried without specifying “-j50”, with “WITHOUT_MANCOMPRESS=y” enabled, and it stopped at:
> 
> make[4]: don't know how to make vi.1. Stop

This is similar to the above in that it has a relative pathname with a subdirectory:

MAN=    ${SRCDIR}/man/vi.1

Most of the existing logic in bsd.man.mk uses :T, but perhaps the rule this commit added
needs to not use :T when defining the target in the !MANSRC case.

The patch below fixes a manual 'make WITHOUT_MANCOMPRESS=yes' in usr.bin/vi in a buildenv
for me and I think would fix the apmd.8 case as well.

diff --git a/share/mk/bsd.man.mk b/share/mk/bsd.man.mk
index 768879b64e60..a2c34eff9e3e 100644
--- a/share/mk/bsd.man.mk
+++ b/share/mk/bsd.man.mk
@@ -169,14 +169,16 @@ ${__target}: ${MANSRC.${__page:T}:U${__page}}
  .endfor
  .else
  .for __page in ${${__group}}
+.if defined(MANSRC.${__page:T})
  .for __target in ${__page:T:S/:/\:/g}
  all-man: ${__target}
-.if defined(MANSRC.${__page:T})
  CLEANFILES+=	${__target}
  ${__target}: ${MANSRC.${__page:T}}
  	${CP} ${.ALLSRC} ${.TARGET}
-.endif
  .endfor
+.else
+all-man: ${__page}
+.endif
  .endfor
  .endif
  .endif

(In a related note, I don't really understand the nested for loops used in much of bsd.man.mk,
once you do 'for page in group' page will already be a single word so the 'for target in page'
will always just be a single iteration IIUC.  Is it just a convenient way to define __target as
a temporary variable?)

-- 
John Baldwin