It turns out that we need a Fontmap file after all, *and* -I. to make
gs find it. Inconsistent results came from stray Fontmap files from
previous debug attempts.
Now generate both fontpath and Fontmap, and hopefully at least one of
them should work. We might, in fact, need both, one for gs to know
where the files are and one for gs to know it is allowed to read them.
The core problem seems to be that gs will find OTF fonts by its normal
discovery mechanisms, but for some reason don't seem to use them
unless it can find them in a Fontmap, Font directory, of CIDFont
directory.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
We need to create a separate paragraph if the help text had used \c
anyway. Putting the enabled/disabled separately for all entries makes
it read a lot cleaner anyway.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
rdsrc.pl requires blank lines around \c paragraph, but warnings.pl
would strip them. Create a *!- prefix to force a blank line.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
config/config.h as generated by autoconf has #undef directives
commented out, but the autoheader-generated template contains them,
and config/unconfig.h should not contain them. Re-introduce
config/config.h.in, and postprocess it to generate config/unconfig.h
by commenting out all the #undef directives, just as configure does.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
<config/unconfig.h> is generated by autoconf (autoheader), but just
like the autoconf/helpers directory contains files for people who
don't have the proper tools, check this file in.
It should even be less annoying change-wise than the helpers...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
\\?\ is supposed to override the hard-coded path limit, but it has
other effects as well, such as not working with relative paths and
paths containing /. On Windows 10 it is possible to set a registry key
to override this option anyway.
Maybe one day we can just use fopen() like on normal systems, even...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: https://bugzilla.nasm.us/show_bug.cgi?id=3392614
Reported-by: Iouri Kharon <bc-info@styx.cabel.net>
The previous code would fail to process any directive if the directive
list was NULL. However, we also need to process the default name
passed to search_pragma_list() (e.g. "elf32"), as well as the global
name (e.g. "output") and call the default handler in that case.
In the process, improve the handling such that if one handler returns
DIRR_UNKNOWN, try calling subsequent handlers in the list.
Finally, factor out as much as possible to generic handlers.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
The inlining of preprocessor tokens has benchmarked about 15% faster in any
realistic scenario than the old code. For an artificial case where
over 50% of symbols are longer than 47 characters there was a slowdown
of about 15%, but that represents an incredibly unlikely scenario in
practice: even if labels are that long (e.g. C++ name mangling) the
majority of tokens should be instructions and registers, which are
uniformly short.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Updating the autoconf helpers unconditionally is very noisy and
probably the wrong thing to do anyway. Check the version numbers and
only update if the current machine has a newer version installed.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Major cleanups of the preprocessor. In particular, the
block-allocation of Token is pretty ridiculous since nearly every
token requires a text allocation anyway. Change the definition of
Token so that only very long tokens (48+ characters on 64-bit systems)
need to be stored out of line.
If malloc() preserves alignment (XXX: glibc doesn't) then this means
that each Token will fit in a cache line.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
paste_tokens() would not null-terminate the buffer before passing it
to tokenize(), resulting in garbage or a memory overwrite.
In several places the next pointers got confused; sometimes causing a
circular list and sometimes an invalid pointer.
Some minor code cleanups while fixing things, too...
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
expand_one_smacro() would corrupt the end of the list if a macro
expanded to another macro with arguments, which was also the last part
of the expansion.
Instead of doing all that testing with ttail, just scan forward at the
end to find the tail pointer; it is O(n) regardless.
Clean up the handling of tokens: use inline functions rather than odd
macros that sometimes modify their arguments and sometimes don't, and
fold some common code into new functions.
The tok_is() and tok_isnt() functions always are used with single
characters, so make it explicitly so (and remove the local hacks used
in some places.)
Allow using nasm_malloc() rather than blocked Tokens; this makes tools
like valgrind more useful in their reports.
For the future, consider making Tokens a separate memory allocation
immediately followed by the text, instead of using a pointer; we
allocate space for the string in almost every case anyway. Also
consider making it a doubly linked list...
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
%ifusable tests to see if a certain %use package is available in this
version of NASM.
%ifusing tests if a certain %use packages is already loaded.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Move the parsing of %use package names to a separate routine, and stop
using get_id() for that purpose -- get_id() is wrong in a number of
ways.
This also means we can drop the error string argument to get_id().
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
While changing this code around to not do redundant lookups, dropped
this NULL pointer check. Oops.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
The "object" symbol keyword would incorrectly generate STT_NOTYPE
instead of STT_OBJECT. Add test for weak object references; they are
different from plain weak references in that they are guaranteed to
resolve to zero if the symbol is not found.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
NASM convention is to use all-upper-case for "real" information, and
mixed-case (upper case common prefix, lower case description) for
meta-information. This is a highly useful distinction.
Thus "LBL_NONE" implies an actual label of type "NONE", as opposed to
no label at all.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Currently, NASM always issues as an unknown symbol any symbol declared
EXTERN. This is highly undesirable when using common header files,
as it might cause the linker to pull in a bunch of unnecessary
modules, depending on how smart the linker is.
Add a new REQUIRED directive which behaves like the old EXTERN, for
the use cases which might still need this behavior.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
A global or extern definition can now contain the keyword "weak" (or
"strong", although that is the default) to create a weak symbol or a
weak external reference, respectively.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Since pp_error_list_macros() was introduced, the only need for
pp_verror() is to suppress error messages in certain contexts. Replace
this function with a preprocessor callback,
preproc->pp_suppress_error(), so we can drop the nasm_verror()
function pointer entirely.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Introduce a new error level, ERR_CRITICAL, beyond which we will
minimize the amount of code that will be executed before we die; in
particular don't execute any memory allocations, and if we somehow end
up recursing, abort() immediately.
Basically, "less than panic, more than fatal."
At this point this level is used by nasm_alloc_failed().
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
If we get a memory allocation failure before preproc is initialized,
we could end up taking a NULL pointer reference while trying to unwind
macros.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Enough users expect the namespace starting with underscore to be safe
for symbols. Change our private namespace from __foo__ to
__?foo?__. Use %defalias to provide backwards compatiblity (by using
%defalias instead of %define, we handle the case properly where the
user changes the value.)
Add a preprocessor directive:
%aliases off
... to disable all smacro aliases and thereby making the namespace
clean.
Finally, fix infinite recursion when seeing %? or %?? due to
paste_tokens(). If we don't paste anything, the expansion is done.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
If we adjust nparams due to default or greedy arguments, we need to
re-terminate the params[] array.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
The smacro argument list cannot be preceded by whitespace, or we
wouldn't be able to define no-argument smacros the expansion of which
starts with (.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Correctly handle empty mmacro arguments that still have preceding
whitespace tokens.
Default mmacro parameters are obtained by count_mmac_params() so they,
too, need to be shifted over by one.
Add an option to list mmacro calls with arguments. Name this -Lm;
remove the old -Lm option to -Ls since it is related to single-line
macros.
Trivially optimize the case where an mmacro is called from within
itself: if all possible mmacros are excluded by loop removal, there is
no need to delve into the mmac processing code.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Fix the (severely broken handling of) varadic macros.
Add a conditional comma operator "%,". This expands to a comma unless
followed by a null expansion of some sort, which allows suppressing
the comma before an empty argument (usually varadic, but not
necessarily.)
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Add a few more types of smacro arguments, and clean stuff up in
process.
If declared with an &, an smacro parameter will be quoted as a string.
If declared with a +, it is a greedy/varadic parameter.
If declared with an !, don't strip whitespace and braces (useful with &).
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
BR 3392603: When doing nested macro definitions, we need %00, %? and
%?? expansion to be deferred to actual expansion time, just as the
other parameters.
Do major cleanups to the mmacro expansion code.
Reported-by: Alexandre Audibert <alexandre.audibert@outlook.fr>
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Ponderance: if data->bits < globalbits, should we actually use
OUT_UNSIGNED rather than OUT_WRAP here?
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
If the address we are using is >= the size of the instruction, then
don't complain on overflow as we can wrap around the top and bottom of
the address space just fine.
Alternatively we could downgrade it to OUT_WRAP in that case.
Reported-by: C. Masloch <pushbx@38.de>
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Don't segfault on a bad %pragma limit. Instead treat a NULL pointer as
an empty string.
Reported-by: Ren Kimura <rkx1209dev@gmail.com>
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
--disable-pedantic passed to configure will now disable
-W, -Wc90-c99-compat, and -pedantic.
Requested-by: Orzkan Sezer <sezeroz@gmail.com>
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
malloc(0) can legitimately return NULL; it does on some systems and
not others. Force the size to 1 byte if the size is 0 coming in,
except for realloc() where this is legitimate and equivalent to
free().
Since this is an abnormal case, and can't even happen with most C
libraries, handle it on the error path, after we already got back a
NULL pointer.
Reported-by: Ozkan Sezer <sezeroz@gmail.com>
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Old versions of gcc didn't define either __GNUC_STDC_INLINE__ or
__GNUC_GNU_INLINE__, but imply the latter.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Separate out function and function pointer attributes, as not all
versions of all compilers support both.
Have macros related to function attributes auto-generated by
autoheader. As a result, rename config.h.in to unconfig.h, to make it
more obvious that it is really intended to be included from some C
programs.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
BR 3392602: mmacros should not nest unless so explicitly specified.
Reported-by: C. Masloch <pushbx@38.de>
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>