My previous tests (with ./configure PERL=false) have been fooled by
configure, that managed to find perl anyway. This time, I ran this on
a Fedora in Docker, without Perl.
* tests/calc.at, tests/diagnostics.at, tests/headers.at,
* tests/input.at, tests/local.at, tests/named-refs.at,
* tests/output.at, tests/regression.at, tests/skeletons.at,
* tests/synclines.at, tests/torture.at: Don't require Perl.
Currently we face test suite failures in different environments,
because of a conflict between the definitions of isnan by gnulib, and
by the C++ library:
262. headers.at:186: testing Sane headers: %locations %debug c++ ...
./headers.at:186: COLUMNS=1000; export COLUMNS; bison --color=no -fno-caret -d -o input.cc input.y
./headers.at:186: $CXX $CXXFLAGS $CPPFLAGS -c -o input.o input.cc
stderr:
In file included from /usr/include/c++/4.8.2/cmath:44:0,
from /usr/include/c++/4.8.2/random:38,
from /usr/include/c++/4.8.2/bits/stl_algo.h:65,
from /usr/include/c++/4.8.2/algorithm:62,
from location.hh:41,
from input.hh:90,
from input.cc:50:
/u/cs/fac/eggert/src/gnu/bison/lib/math.h: In function 'bool isnan(double)':
/u/cs/fac/eggert/src/gnu/bison/lib/math.h:2849:1: error: new declaration 'bool isnan(double)'
_GL_MATH_CXX_REAL_FLOATING_DECL_2 (isnan, isnan, bool)
^
In file included from /usr/include/features.h:375:0,
from /usr/include/c++/4.8.2/x86_64-redhat-linux/bits/os_defines.h:39,
from /usr/include/c++/4.8.2/x86_64-redhat-linux/bits/c++config.h:2097,
from /usr/include/c++/4.8.2/cstdlib:41,
from input.hh:48,
from input.cc:50:
/usr/include/bits/mathcalls.h:235:1: error: ambiguates old declaration 'int isnan(double)'
__MATHDECL_1 (int,isnan,, (_Mdouble_ __value)) __attribute__ ((__const__));
^
There might be something to do in gnulib about this, but I believe
that gnulib should not be used in the test suite in the first place.
The test suite should work with other compilers than the one used to
compile the package. For a start, Bison sources are more
demanding (C99) than the generated parsers. Last time I tried, tcc
for example, was not able to compile Bison, yet our generated parsers
should compile cleanly with it.
Besides the problem at hand is with the C++ compiler, with is not the
one used to set up gnulib at configuration-time (config.h is mainly
built from probing the C compiler).
We should really not depend on gnulib in tests.
This was introduced in 2001 to check whether including
stdlib.h/string.h is safe thanks to STDC_HEADERS
(2ce1014469). Today, we assume at least
a C90 compiler, it should be safe enough.
* tests/local.at, tests/testsuite.h: Do not include config.h.
* tests/atlocal.in (conftest.cc): Likewise.
(CPPFLAGS): Do not expose lib/, as because of this we might picked up
gnulib replacement headers for system headers.
* tests/input.at: Use int instead of ptrdiff_t, for easier portability
(some machine on the CI did not find ptrdiff_t).
* tests/c++.at: Add missing include for getchar.
A number of portability issues with GCC 4.6 .. 4.9 (inclusive):
input.c:184:7: error: "UCHAR_MAX" is not defined [-Werror=undef]
#elif UCHAR_MAX <= INT_MAX
^
input.c:184:20: error: "INT_MAX" is not defined [-Werror=undef]
#elif UCHAR_MAX <= INT_MAX
^
input.c:202:7: error: "USHRT_MAX" is not defined [-Werror=undef]
#elif USHRT_MAX <= INT_MAX
^
input.c:202:20: error: "INT_MAX" is not defined [-Werror=undef]
#elif USHRT_MAX <= INT_MAX
^
* data/skeletons/c.m4 (b4_c99_int_type_define): Don't assume they are
defined.
Flex should not be required to build Bison or run the test suite (of
course it is needed for maintaining Bison). Yet the Automake
conditional FLEX_WORKS does not work.
* m4/flex.m4 (_AC_PROG_LEX_YYTEXT_DECL): Since this is called
conditionally, don't define LEX_IS_FLEX here, but rather...
(AC_PROG_LEX): here.
* configure.ac: Be more cautious about possibly undefined variables.
Because the checking of the grammar is made by phases after the whole
grammar was read, we sometimes have diagnostics that look weird. In
some case, within one type of checking, the entities are not checked
in the order in which they appear in the file. For instance, checking
symbols is done on the list of symbols sorted by tag:
foo.y:1.20-22: warning: symbol BAR is used, but is not defined as a token and has no rules [-Wother]
1 | %destructor {} QUX BAR
| ^~~
foo.y:1.16-18: warning: symbol QUX is used, but is not defined as a token and has no rules [-Wother]
1 | %destructor {} QUX BAR
| ^~~
Let's sort them by location instead:
foo.y:1.16-18: warning: symbol 'QUX' is used, but is not defined as a token and has no rules [-Wother]
1 | %destructor {} QUX BAR
| ^~~
foo.y:1.20-22: warning: symbol 'BAR' is used, but is not defined as a token and has no rules [-Wother]
1 | %destructor {} QUX BAR
| ^~~
* src/location.h (location_cmp): Be robust to empty file names.
* src/symtab.c (symbol_cmp): Sort by location.
* tests/input.at: Adjust expectations.
From
input.y:1.17-19: warning: symbol baz is used, but is not defined as a token and has no rules [-Wother]
1 | %printer {} foo baz
| ^~~
to
input.y:1.17-19: warning: symbol 'baz' is used, but is not defined as a token and has no rules; did you mean 'bar'? [-Wother]
1 | %printer {} foo baz
| ^~~
| bar
* bootstrap.conf: We need fstrcmp.
* src/symtab.c (symbol_from_uniqstr_fuzzy): New.
(complain_symbol_undeclared): Use it.
* tests/diagnostics.at (Suggestions): New.
* data/bison-default.css (insertion): Rename as...
(fixit-insert): this, as this is what GCC uses.
* src/symtab.c (complain_symbol_undeclared): New.
Use it.
Use quote on the guilty symbol (like GCC does, and we also do
elsewhere).
* tests/input.at: Adjust.
Both are stored in a hash, and back in the days, we used to iterate
over these tables using hash_do_for_each. However, the order of
traversal was not deterministic, which was a nuisance for
deterministic output (and therefore also a problem for tests). So at
some point (83b60c97ee) we generated a
sorted list of these symbols, and symbols_do actually iterated on that
list. But we kept the constraints of using hash_do_for_each, which
requires a lot of ceremonial code, and makes it hard/unnatural to
preserve data between iterations (see the next commit).
Alas, this is C, not C++.
Let's remove this abstraction, and directly iterate on the sorted
tables.
* src/symtab.c (symbols_do): Remove.
Adjust callers to use a simple for-loop instead.
(table_sort): New.
(symbols_check_defined): Use it.
(symbol_check_defined_processor, symbol_pack_processor)
(semantic_type_check_defined_processor, symbol_translation_processor):
Remove.
Simplify the corresponding functions (that no longer need to return a
bool).
This commit adds the suggestion in green, on the line below the
caret-and-tildes.
foo.y:1.1-14: warning: deprecated directive: '%error-verbose', use '%define parse.error verbose' [-Wdeprecated]
1 | %error-verbose
| ^~~~~~~~~~~~~~
| %define parse.error verbose
The current approach, with location_caret_suggestion, is fragile:
there's a protocol of calls to the complain functions which is strict.
We should rather have a richer structure describing the diagnostics,
including with submessages such as the suggestions, passed in the end
to the routines in charge of formatting and printing them.
* src/location.h, src/location.c (location_caret_suggestion): New.
* src/complain.c (deprecated_directive): Use it.
* tests/diagnostics.at, tests/input.at: Adjust expectations.
* src/location.c (caret_set_file): New.
Store the current line's length in caret_info.line_len.
Pay attention to fseek's return value.
Extracted from...
(location_caret): here.
I no longer agree with that item, there are indeed two things to
report: lack of definition, and being useless. We could have either
one without the other, they are not directly related.
input.c: In function 'int yyparse()':
input.c: error: conversion to 'long int' from 'long unsigned int'
may change the sign of the result [-Werror=sign-conversion]
yyes_capacity = sizeof yyesa / sizeof *yyes;
^
cc1plus: all warnings being treated as errors
* data/skeletons/yacc.c: here.
GCC 4.8 reports:
input.y:57:33: error: conversion to 'int' from 'long unsigned int'
may alter its value [-Werror=conversion]
int input_elts = sizeof input / sizeof input[0];
^
* tests/local.at (AT_YYLEX_DEFINE(c)): Add a cast (sorry, Paul!).
For instance with GCC 4.9 and --enable-gcc-warnings:
25. input.at:1201: testing Torturing the Scanner ...
../../tests/input.at:1344: $CC $CFLAGS $CPPFLAGS -c -o input.o input.c
stderr:
input.c:239:18: error: "__STDC_VERSION__" is not defined [-Werror=undef]
# elif 199901 <= __STDC_VERSION__
^
input.c:256:18: error: "__STDC_VERSION__" is not defined [-Werror=undef]
# elif 199901 <= __STDC_VERSION__
^
* data/skeletons/yacc.c: Check that __STDC_VERSION__ is defined before
using it.
On the CI with GCC 6:
examples/c++/calc++/parser.cc:845:5: error: 'ptrdiff_t' was not declared in this scope
ptrdiff_t yycount = 0;
^~~~~~~~~
examples/c++/calc++/parser.cc:845:5: note: suggested alternatives:
/usr/include/x86_64-linux-gnu/c++/6/bits/c++config.h:202:28: note: 'std::ptrdiff_t'
typedef __PTRDIFF_TYPE__ ptrdiff_t;
^~~~~~~~~
* data/skeletons/lalr1.cc: Qualify ptrdiff_t and size_t with std::.
Currently we properly use the "best" integral type for tables,
including those storing state numbers. However the variables for
state numbers used in yyparse (and its dependencies such as
yy_stack_print) still use int16_t invariably. As a consequence, very
large models overflow these variables.
Let's use the "best" type for these variables too. It turns out that
we can still use 16 bits for twice larger automata: stick to unsigned
types.
However using 'unsigned' when 16 bits are not enough is troublesome
and generates tons of warnings about signedness issues. Instead,
let's use 'int'.
Reported by Tom Kramer.
https://lists.gnu.org/archive/html/bug-bison/2019-09/msg00018.html
* data/skeletons/yacc.c (b4_state_num_type): New.
(yy_state_num): Be computed from YYNSTATES.
* tests/linear: New.
* tests/torture.at (State number type): New.
Use it.
Instead of
#define YYPACT_NINF -130
#define yypact_value_is_default(Yystate) \
(!!((Yystate) == (-130)))
generate
#define YYPACT_NINF (-130)
#define yypact_value_is_default(Yyn) \
((Yyn) == YYPACT_NINF)
* data/skeletons/c.m4 (b4_table_value_equals): Add support for $4.
* data/skeletons/glr.c, data/skeletons/yacc.c: Use it.
Also, use shorter macro argument names, the name of the macro is clear
enough.
* upstream/maint:
c++: add copy ctors for compatibility with the IAR compiler
CI: show git status
CI: disable ICC
tests: pass -jN from Make to the test suite
quotearg: avoid leaks
maint: post-release administrivia
Reported by Andreas Damm.
https://savannah.gnu.org/support/?110032
* data/skeletons/lalr1.cc (stack_symbol_type::operator=): New
overload, const, to please the IAR C++ compiler (version ca 2013).
* bootstrap.conf: We need winsz-ioctl and winsz-termios.
* src/location.c (columns): Use winsize to get the number of
columns.
Code taken from the GNU Coreutils.
* src/location.h, src/location.c (caret_init): New.
* src/complain.c (complain_init): Call it.
* tests/bison.in: Export COLUMNS so that users of tests/bison can
enjoy proper line truncation.
* src/location.c (min_int, columns): New.
(location_caret): Compute the line width. Based on it, compute how
many columns must be skipped before the quoted location and truncated
after, to fit the sceen width.
* tests/local.at (AT_QUELL_VALGRIND): Transform into...
(AT_SET_ENV_IF, AT_SET_ENV): these.
Define COLUMNS to protect the test suite from the user's environment.
So far diagnostics were cheating: in addition to the 'column' field of
locations (based on actual screen width per multibyte characters and
on tabulation expansion), the scanner sets the 'byte' field.
Diagnostics used this byte count to decide where to insert (color)
style.
We want to be able to truncate the quoted lines when there are too
wide to fit the screen. This requires that the diagnostics learn how
to count columns, the byte-in-boundary trick no longer works.
Bytes are still used for fix-its.
* bootstrap.conf: We need mbfile for mbf_getc.
* src/location.c (caret_info): We need an mbfile.
(caret_set_file): Initialize it.
(caret_getc): Convert to mbfile.
(location_caret): Instead of relying on the byte position to decide
where to insert the color style, count the current column using
boundary_compute.
* src/location.c (caret_info): Replace file and line with pos, a
boundary. This will allow us to use features of the boundary type,
such as boundary_compute.
The handling of the contributions of the tabulations in the columns is
burried inside location_compute. We will soon be willing to use the
boundary part of the computation (to compute the current column number
each time we read a multibyte char).
* src/location.c (boundary_compute): New, extracted from...
(location_compute): here.
This changes the traces from
Reading a token:
Now at end of input.
to
Reading a token:
Next token is token $end (7FFEE56E6474)
which is ok. Actually it is even better, as it gives the location
when locations are enabled, and is clearer when rules explicitly use
the EOF token.
* data/skeletons/lalr1.d (yytranslate_): Handle eof here, as is done
in lalr1.cc.
Bison used to feature %raw, documented as follows:
@item %raw
The output file @file{@var{name}.h} normally defines the tokens with
Yacc-compatible token numbers. If this option is specified, the
internal Bison numbers are used instead. (Yacc-compatible numbers start
at 257 except for single character tokens; Bison assigns token numbers
sequentially for all tokens starting at 3.)
Unfortunately, as far as I can tell, it never worked: token numbers
are indeed changed in the generated tables (from external token number
to internal), yet the code was still applying the mapping from
external token numbers to internal token numbers.
This commit reintroduces the feature as it was expected to be.
* data/skeletons/bison.m4 (b4_token_format): When api.token.raw is
enabled, use the internal token number.
* data/skeletons/yacc.c (yytranslate): Don't emit if api.token.raw is
enabled.
(YYTRANSLATE): Adjust.
We used to treat lone CRs (\r, aka ^M) as regular NLs (\n), probably
to please Classic MacOS. As of today, it makes more sense to treat \r
like a plain white space character.
https://lists.gnu.org/archive/html/bison-patches/2019-09/msg00027.html
* src/scan-gram.l (no_cr_read): Remove. Instead, use...
(eol): this new abbreviation denoting end-of-line.
* src/location.c (caret_getc): New.
(location_caret): Use it.
* tests/diagnostics.at (Carriage return): Adjust expectations.
(CR NL): New.
When the input file contains lone CRs (aka, ^M, \r), the locations see
a new line. Diagnostics look only at \n as end-of-line, so sometimes
there is an offset in diagnostics. Worse yet: sometimes we loop
endlessly waiting for \n to come from a continuous stream of EOF.
Fix that:
- check for EOF
- beware not to call end_use_class if begin_use_class was not
called (which would abort). This could happen if the actual
line is shorter that the expected one.
Prompted by a (private) report from Marc Schönefeld.
* src/location.c (location_caret): here.
* tests/diagnostics.at (Carriage return): New.
When the input file contains lone CRs (aka, ^M, \r), the locations see
a new line. Diagnostics look only at \n as end-of-line, so sometimes
there is an offset in diagnostics. Worse yet: sometimes we loop
endlessly waiting for \n to come from a continuous stream of EOF.
Fix that:
- check for EOF
- beware not to call end_use_class if begin_use_class was not
called (which would abort). This could happen if the actual
line is shorter that the expected one.
Prompted by a (private) report from Marc Schönefeld.
* src/location.c (location_caret): here.
* tests/diagnostics.at (Carriage return): New.
With
%token EOF 0 EOF 0
we get
input.y:3.14-16: warning: symbol EOF redeclared [-Wother]
3 | %token EOF 0 EOF 0
| ^~~
input.y:3.8-10: previous declaration
3 | %token EOF 0 EOF 0
| ^~~
Assertion failed: (nsyms == ntokens + nvars), function check_and_convert_grammar,
file /Users/akim/src/gnu/bison/src/reader.c, line 839.
Reported by Marc Schönefeld.
* src/symtab.c (symbol_user_token_number_set): Register only the
first definition of the end of input token.
* tests/input.at (Symbol redeclared): Check that case.
The CI, with CC='gcc-7 -fsanitize=undefined,address
-fno-omit-frame-pointer', reports:
calc.cc:1652:50: runtime error: load of value 190, which is not a valid value for type 'bool'
../../tests/calc.at:867: cat stderr
--- expout 2019-09-05 20:30:37.887257545 +0000
+++ /home/travis/build/bison-3.4.1.72-79a1-dirty/_build/tests/testsuite.dir/at-groups/438/stdout 2019-09-05 20:30:37.887257545 +0000
@@ -1 +1,2 @@
syntax error
+calc.cc:1652:50: runtime error: load of value 190, which is not a valid value for type 'bool'
438. calc.at:867: 438. Calculator glr.cc (calc.at:867): FAILED (calc.at:867)
The problem is that yylookaheadNeeds is not initialized in
yyinitStateSet, and when it is copied, the value is not 0 or 1.
* data/skeletons/glr.c (yylookaheadNeeds): Initialize yylookaheadNeeds.
hash_initialize returns NULL when out of memory. Check for it, and
die cleanly instead of crashing.
Reported by 江 祖铭 (Zu-Ming Jiang).
https://lists.gnu.org/archive/html/bug-bison/2019-08/msg00015.html
* src/muscle-tab.c, src/state.c, src/symtab.c, src/uniqstr.c:
Check the value returned by hash_initialize.
* .travis.yml: Bionic is now available, with GCC8.
GCC7 sanitizers work, but they are too longer: cover only part 1.
Redefine part 1 and part 2 so that part 1 is really the core of the
tests: not playing with POSIX and C++ compiler for C code.
Akim Demaille