GNU bug report logs - #75964
Switching the Emacs build to -Wswitch-enum in src/

Previous Next

Full log

View this message in rfc822 format

From: Paul Eggert <eggert <at> cs.ucla.edu>
To: Pip Cet <pipcet <at> protonmail.com>, 75964 <at> debbugs.gnu.org, Stefan Kangas <stefankangas <at> gmail.com>
Subject: bug#75964: Switching the Emacs build to -Wswitch-enum in src/
Date: Fri, 31 Jan 2025 13:46:53 -0800

On 2025-01-31 03:05, Pip Cet wrote:
> This GCC version is also patched to "infer" a switch type

Suppose the inference is wrong: how does one easily pacify the modified 
GCC? Is it something like "case +PVEC_NORMAL:"?

Suppose one case has an enum type and another case has some other type: 
is that an error?

> +    case PVEC_FREE: eassume (0);
> +    default: eassume (0);

Must this be written that way? Suppose I do this:

   case PVEC_FREE:
   default:
     eassume (false);

Does that also pacify the modified GCC? Also, does emacs_abort () also 
pacify the modified GCC?

Perhaps we should define a new macro eunreachable () that expands to 
eassume (false). eunreachable : unreachable :: eassert : assert. This 
might be a good idea regardless of anything else we do.


> alloc.c was a simple case because the enums are all internal
> to Emacs and unlikely to change drastically without people touching the
> code anyway.  image.c is very different in that regard.

How does that sort of thing pan out?


> +  ws="$ws -Wswitch-enum"

This won't work with the latest stable GCC, as it will cause false 
positives. We need a way to have Emacs use -Wswitch-enum only with the 
proposed patch installed. Perhaps the new GCC behavior should be enabled 
by a new GCC option (not -Wswitch-enum), so that it's easy to test 
whether the proposed GCC behavior is available.


> +ALL_CFLAGS = ${BASE_CFLAGS} ${PROFILING_CFLAGS} ${LDFLAGS} ${CPPFLAGS} ${CFLAGS} ${PACIFYING_CFLAGS}
>  CPP_CFLAGS = ${BASE_CFLAGS} ${PROFILING_CFLAGS} ${CPPFLAGS} ${CFLAGS}
>  o
>  ALL_CXXFLAGS = $(filter-out ${NON_CXX_CFLAGS},${BASE_CFLAGS}) \
>    ${PROFILING_CFLAGS} ${LDFLAGS} ${CPPFLAGS} ${CFLAGS} ${CXXFLAGS} ${HAIKU_CFLAGS}
>  
> +unfixed_Wswitch_obj += ctags${EXEEXT}
> +unfixed_Wswitch_obj += etags${EXEEXT}

This sort of thing can be confusing. Do we need it at all? That is, once 
the fixes are in the C source files we shouldn't need to change the 
Makefile.

Alternatively, perhaps it'd be better to use a pragma in the (I hope) 
rare places we need to pacify GCC. E.g., how we currently treat 
-Wanalyzer-null-dereference in four .c files.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.