git: 2a50a9de8340 - main - terminfo: add terminfo database

[Yuri Pankov]

Cy Schubert wrote:
> In message <202102251327.11PDR4eC083842 at gitrepo.freebsd.org>, Baptiste 
> Daroussi
> n writes:
>> The branch main has been updated by bapt:
>>
>> URL: https://cgit.FreeBSD.org/src/commit/?id=2a50a9de8340f08bd876e9e5993332ae
>> 14376f80
>>
>> commit 2a50a9de8340f08bd876e9e5993332ae14376f80
>> Author:     Baptiste Daroussin <bapt at FreeBSD.org>
>> AuthorDate: 2021-02-23 16:17:32 +0000
>> Commit:     Baptiste Daroussin <bapt at FreeBSD.org>
>> CommitDate: 2021-02-25 13:25:32 +0000
>>
>>     terminfo: add terminfo database
>>     
>>     Tested by:      manu, jbeich
>> ---
>>  share/Makefile          |  1 +
>>  share/terminfo/Makefile | 34 ++++++++++++++++++++++++++++++++++
>>  2 files changed, 35 insertions(+)
>>
>> diff --git a/share/Makefile b/share/Makefile
>> index c4e12b05f7db..d6854b230ae5 100644
>> --- a/share/Makefile
>> +++ b/share/Makefile
>> @@ -26,6 +26,7 @@ SUBDIR=	${_colldef} \
>>  	${_syscons} \
>>  	tabset \
>>  	termcap \
>> +	terminfo \
>>  	${_timedef} \
>>  	${_vt} \
>>  	${_zoneinfo}
>> diff --git a/share/terminfo/Makefile b/share/terminfo/Makefile
>> new file mode 100644
>> index 000000000000..7bb11f3fdf24
>> --- /dev/null
>> +++ b/share/terminfo/Makefile
>> @@ -0,0 +1,34 @@
>> +PACKAGE=	runtime
>> +
>> +.PATH: ${SRCTOP}/contrib/ncurses/misc
>> +TINFOBUILDDIR=	${.OBJDIR}/builddir
>> +CLEANDIRS+=	builddir
>> +
>> +.include <src.tools.mk>
>> +
>> +.if !defined(_SKIP_BUILD)
>> +all: terminfo
>> +.endif
>> +META_TARGETS+=	terminfo install-terminfo
>> +
>> +terminfo: terminfo.src
>> +	mkdir -p ${TINFOBUILDDIR}
>> +	${TIC_CMD} -x -o ${TINFOBUILDDIR} ${.ALLSRC}
>> +
>> +.if make(*install*)
>> +TINFOS!=	 cd ${TINFOBUILDDIR} && find * -type f | LC_ALL=C sort
>> +TINFOSDIRS=	${TINFOS:C/(.).*/\1/g:O:u}
>> +.endif
>> +
>> +beforeinstall: install-terminfo
>> +install-terminfo:
>> +	mkdir -p ${DESTDIR}/usr/share/terminfo
>> +	cd ${DESTDIR}/usr/share/terminfo; mkdir -p ${TINFOSDIRS}
>> +.for f in ${TINFOS}
>> +	${INSTALL} ${TAG_ARGS} \
>> +	    -o ${BINOWN} -g ${BINGRP} -m ${NOBINMODE} \
>> +	    ${TINFOBUILDDIR}/${f} ${DESTDIR}/usr/share/terminfo/${f}
>> +.endfor
>> +
>> +.include <bsd.prog.mk>
>> +
>>
> 
> I think this had some unintended consequences, a POLA violation that should 
> be documented. Our termcap didn't support alternate screen buffers while 
> terminfo does. Termcap did through t_te and t_ti but we didn't use them. 
> Terminfo as delivered from the ncurses factory does supply rmcup and smcup; 
> our fullscreen apps such as nvi and top see them and use the alternate 
> screen buffer.

I don't think this is correct, it's 'ti' and 'te' for termcap and we use
those a lot in our database, e.g. xterm-clear or even default screen and
tmux entries, so it's always there for me being tmux user.

> The use of the alternate screen buffer is one of the things I hate about 
> Linux. After running vi or top I expect the output to remain on the screen 
> so I can use some bit of information in my next command. We now have this 
> behaviour too.

It's matter of taste, yes, and I like it.

> The alternate screen buffer was added to xterm an ice age ago. Most modern 
> terminal emulators support it. As much as I feel the need to yank out rmcup 
> an smcup from our terminfo database, we are using a feature, as distasteful 
> as it may feel. This change probably needs a relnotes=yes and an UPDATING 
> entry advising users that they can either change $TERM to xterm1, vt100, or 
> some other terminal definition that doesn't support rmcup and smcup, or 
> learn to like it.
> 
> OTOH, maybe this is something we as a community don't really want and we 
> remove the rmcup and smcup definitions from our terminfo database. 
> Personally, I believe this is progress -- progress I'm not particularly 
> enamoured with but progress nonetheless. But I think we need to discuss, 
> come to some sort of decision and document it.

Removing those entirely will break POLA which started this discussion;
to not break POLA we should rather "fix" the xterm entry to exclude
smcup/rmcup to be the same as termcap equivalent.