bug-ncurses
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 20190630 Breaks fallback.c Generation


From: Thomas Dickey
Subject: Re: 20190630 Breaks fallback.c Generation
Date: Sat, 13 Jul 2019 13:56:54 -0400
User-agent: NeoMutt/20170113 (1.7.2)

On Tue, Jul 09, 2019 at 09:11:38PM -0400, Thomas Dickey wrote:
> On Wed, Jul 10, 2019 at 12:50:35AM +0000, Connor Nolan wrote:
> > ./configure \
> >     --host=$1 \
> >     ac_cv_header_locale_h=no \
> >     --disable-stripping \
> >     --enable-const \
> >     --enable-ext-colors \
> >     --enable-ext-mouse \
> >     --enable-overwrite \
> >     --enable-widec \
> >     --without-ada \
> >     --without-cxx-binding \
> >     --without-debug \
> >     --without-tests \
> >     --with-fallbacks='dumb,xterm-256color' \
> ...
> > clang: warning: argument unused during compilation: '--param 
> > max-inline-insns-single=1200' [-Wunused-command-line-argument]
> > "../misc/terminfo.src", line 4766, terminal 'xterm-16color': error writing 
> > /builds/TheBrokenRail/letscode/scripts/ncurses/ncurses/tmp_info/x/xterm-16color
> 
> ...
> 
> That looks like the same issue that I ran into (the older version of
> tic/infocmp can't handle the newer terminfo.src).  I solved it by
> changing my build to use a newer tic/infocmp (in this case, the
> most recent one, though in my earlier workaround on 20190609, "6.0"
> was good enough).

I spent some time today comparing the installed terminfo versus the program
versions.  Basically what happened was that there was a bug due to reusing
the same terminal description when evaluating the "use=" clauses and merging
them.

I fixed that in

20190609
        + modify _nc_merge_entry() to make a copy of the data which it merges,
          to avoid modifying the source-data when aligning extended names.

Sometimes (depending on the contents of terminfo.src), the program
died.  Usually it didn't, but adding or removing a "use=" clause from
an entry could have triggered the bad behavior.

Even when it didn't make tic/infocmp die, the reused data could be wrong
(since it reflects a merge of entries rather than the expected building
block).  That was why I commented out some of the extensions here:

        + add mintty, mintty-direct (adapted from patch by Thomas Wolff).
          Some of the suggested user-defined capabilities are commented-out,
          to allow builds with ncurses 5.9 and 6.0

-- 
Thomas E. Dickey <address@hidden>
https://invisible-island.net
ftp://ftp.invisible-island.net

Attachment: signature.asc
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]