This fixes more differences compared to "gcc -E" so removes several
cases of erroneously failing test cases. The implementation isn't very
elegant, but it is functional.
We fix this by moving printing up to the top-level "input" action and
tracking whether a space is needed between one token and the next.
This fixes all actual bugs in test-suite output, but does leave some
tests failing due to differences in the amount of whitespace produced,
(which aren't actual bugs per se).
This whitespace was not part of anything being tested, and it
introduces differences (that we don't actually care about) between the
output of "gcc -E" and glcpp.
Just eliminate this extra whitespace to reduce spurious test-case
failures.
Sometime back the output of glcpp started differing from the output of
"gcc -E" in the amount of whitespace in emitted. At the time, I
switched the test suite to use "diff -w" to ignore this. This was a
mistake since it ignores whitespace entirely. (I meant to use "diff
-b" which ignores only changes in the amount of whitespace.)
So bugs have since been introduced that the test suite doesn't
notice. For example, glcpp is producing "twotokens" where it should be
producing "two tokens".
Let's stop ignoring whitespace in the test suite, which currently
introduces lots of failures---some real and some spurious.
The fix here is quite simple (and actually only deletes code). When
expanding a macro, we don't return a ',' as a unique token type, but
simply let it fall through to the generic case.
The specification says that commas within a parenthesized group,
(that's not a function-like macro invocation), are passed through
literally and not considered argument separators in any outer macro
invocation.
Add support and a test for this case. This support makes a third
occurrence of the same "FUNC_MACRO (" shift/reduce conflict appear, so
expect that.
This change does introduce a fairly large copy/paste block in the
grammar which is unfortunate. Perhaps if I were more clever I'd find a
way to share the common pieces between argument and argument_or_comma.
The specification of the preprocessor in C99 says that when we see a
macro name that we are already expanding that we refuse to expand it
now, (which we've done for a while), but also that we refuse to ever
expand it later if seen in other contexts at which it would be
legitimate to expand.
We add a test case for that here, and fix it to work. The fix takes
advantage of a new token_t value for tokens and argument words along
with the recently added IDENTIFIER_FINALIZED token type which
instructs the parser to not even look for another expansion.
There's not yet any change in functionality here, (at least according
to the test suite). But we now have the option of specifying a type
for each string in the token list. This will allow us to finalize an
unexpanded macro name so that it won't be subjected to excess
expansion later.
Previously, we would pass original strings back to the original lexer
whenever we needed to re-lex something, (such as an expanded macro or
a macro argument). Now, we instead parse the macro or argument
originally to a string list, and then re-lex by simply returning each
string from this list in turn.
We do this in the recently added glcpp_parser_lex function that sits
on top of the lower-level glcpp_lex that only deals with text.
This doesn't change any behavior (at least according to the existing
test suite which all still passes) but it brings us much closer to
being able to "finalize" an unexpanded macro as required by the
specification.
We rename the generated lexer from yylex to glcpp_lex. Then we
implement our own yylex function in glcpp-parse.y that calls
glcpp_lex. This doesn't change the behavior at all yet, but gives us a
place where we can do implement alternate lexing in the future.
(We want this because instead of re-lexing from strings for macro
expansion, we want to lex from pre-parsed token lists. We need this so
that when we terminate recursion due to an already active macro
expansion, we can ensure that that symbol never gets expanded again
later.)
The support for an object-like amcro within a macro-invocation
argument was also implemented at one level too high in the
grammar. Fortunately, this is a very simple fix.
The previous fix added FUNC_MACRO to a production one higher in teh
grammar than it should have. So it prevented a FUNC_MACRO from
appearing as part of a mutli-token argument rather than just alone as
an argument. Fix this (and add a test).
This adds a second shift/reduce conflict to our grammar. It's basically the
same conflict we had previously, (deciding to shift a '(' after a FUNC_MACRO)
but this time in the "argument" context rather than the "content" context.
It would be nice to not have these, but I think they are unavoidable
(withotu a lot of pain at least) given the preprocessor specification.
This case worked previously, but broke in the recent rewrite of
function- like macro expansion. The recursion was still terminated
correctly, but any parenthesized expression after the macro name was
still being swallowed even though the identifier was not being
expanded as a macro.
The fix is to notice earlier that the identifier is an
already-expanding macro. We let the lexer know this through the
classify_token function so that an already-expanding macro is lexed as
an identifier, not a FUNC_MACRO.
The rewrite her discards the functions that did direct, recursive
expansion of macro values. Instead, the parser now pushes the macro
definition string over to a stack of buffers for the lexer. This way,
macro expansion gets access to all parsing machinery.
This isn't a small change, but the result is simpler than before (I
think). It passes the entire test suite, including the four tests
added with the previous commit that were failing before.
Many of these look quite similar to existing tests that are handled
correctly, yet none of these work. For example, in test 30 we have a
simple non-function macro "foo" that is defined as "bar(baz(success))"
and obviously non-function macro expansion has been working for a long
time. Similarly, if we had text of "bar(baz(success))" it would be
expanded correctly as well.
But when this otherwise functioning text appears as the body of a
macro, things don't work at all.
This is pointing out a fundamental problem with the current
approach. The current code does a recursive expansion of a macro
definition, but this doesn't involve the parsing machinery, so it
can't actually handle things like an arbitrary nesting of parentheses.
The fix will require the parser to stuff macro values back into the
lexer to get at all of the existing machinery when expanding macros.
The test has a newline before the left parenthesis, and newlines to
separate the parentheses from the argument.
The fix involves more state in the lexer to only return a NEWLINE
token when termniating a directive. This is very similar to our
previous fix with extra lexer state to only return the SPACE token
when it would be significant for the parser.
With this change, the exact number and positioning of newlines in the
output is now different compared to "gcc -E" so we add a -B option to
diff when testing to ignore that.
The most recent fix to the parser introduced a shift/reduce
conflict. We document this conflict here, and tell bison that it need
not report it (since I verified that it's being resolved in the
direction desired).
For the record, I did write additional lexer code to eliminate this
conflict, but it was quite fragile, (would not accept a newline
between a function-like macro name and the left parenthesis, for
example).
This has the added advantage that it will stop traversing the tree as
soon as the first call is found.
The output of all test cases was verified to be the same using diff.
That is, when a function-like macro appears in the content without
parentheses it should be accepted and passed on through, (previously
the parser was regarding this as a syntax error).
The test case here is simply "#define foo foo" and "#define bar foo"
and then attempting to expand "bar".
Previously, our termination condition for the recursion was overly
simple---just looking for the single identifier that began the
expansion. We now fix this to maintain a stack of identifiers and
terminate when any one of them occurs in the replacement list.
The first bug was not allowing whitespace between '#' and the
directive name.
The second bug was swallowing a terminating newline along with any
trailing whitespace on a line.
With these two fixes, and the previous commit to stop emitting SPACE
tokens, the recently added extra-whitespace test now passes.
This reverts the unconditional return of SPACE tokens from the lexer
from commit 48b94da099 .
That commit seemed useful because it kept the lexer simpler, but the
presence of SPACE tokens is causing lots of extra complication for the
parser itself, (redundant productions other than whitespace
differences, several productions buggy in the case of extra
whitespace, etc.)
Of course, we'd prefer to never have any whitespace token, but that's
not possible with the need to distinguish between "#define foo()" and
"#define foo ()". So we'll accept a little bit of pain in the lexer,
(enough state to support this special-case token), in exchange for
keeping most of the parser blissffully ignorant of whether tokens are
separated by whitespace or not.
This change does mean that our output now differs from that of "gcc -E",
but only in whitespace. So we test with "diff -w now to ignore those
differences.
We were correctly parsing this already, but simply not returning any
value (for no good reason). Fortunately the fix is quite simple.
This makes the test added in the previous commit now pass.
The macro invocation is defined to consume all text between a set of
matched parentheses. We previously tested for inner parentheses from a
nested function-like macro invocation. Here we test for inner
parentheses occuring on their own, (not part of another macro
invocation).
