https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120610
Bug ID: 120610
Summary: pp_token_lists are not always balanced
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: other
Assignee: unassigned at gcc dot gnu.org
Reporter: dmalcolm at gcc dot gnu.org
Blocks: 116792
Target Milestone: ---
In the experimental-html output sink, html_token_printer::print_tokens takes a
pp_token_list and prints it as html.
Unfortunately this sometimes doesn't fully work; I've seen:
./xgcc -B. -S -fanalyzer -fdiagnostics-add-output=experimental-html
../../src/gcc/testsuite/gcc.dg/bad-binary-ops.c
where the BEGIN_QUOTE END_QUOTE and highlighting aren't balanced, with "aka"
The token list in question is:
TEXT("invalid operands to binary / (have "), BEGIN_QUOTE,
BEGIN_COLOR("highlight-a"), TEXT("__m128’ {aka ‘float’}"), END_COLOR, TEXT("
and "), BEGIN_QUOTE, BEGIN_COLOR("highlight-b"), TEXT("const int *"),
END_COLOR, END_QUOTE, TEXT(")")]
which hierarchically is:
TEXT("invalid operands to binary / (have "),
BEGIN_QUOTE,
BEGIN_COLOR("highlight-a"),
TEXT("__m128’ {aka ‘float’}"), <--- note there's no END_QUOTE here to
match the BEGIN_QUOTE
END_COLOR,
TEXT(" and "),
BEGIN_QUOTE,
BEGIN_COLOR("highlight-b"),
TEXT("const int *"),
END_COLOR,
END_QUOTE,
TEXT(")")]
(unclosed END_QUOTE)
and it's probably should instead be:
TEXT("invalid operands to binary / (have "),
BEGIN_QUOTE,
BEGIN_COLOR("highlight-a"),
TEXT("__m128")
END_COLOR
END_QUOTE <--- here's the missing END_QUOTE, and more fine-grained nodes for
the rest of the type
TEXT (" {aka ")
BEGIN_QUOTE
BEGIN_COLOR("highlight-a"),
TEXT("float")
END_COLOR,
END_QUOTE
TEXT("} and "),
BEGIN_QUOTE,
BEGIN_COLOR("highlight-b"),
TEXT("const int *"),
END_COLOR,
END_QUOTE,
TEXT(")")]
which is properly nested.
This leads to not-quite-correct generated HTML.
Something's going wrong inside the "aka" handling.
For now I'm going to firewall the html_token_printer so it has its own
xml::printer to restrict the scope of the mismatches.
Ultimately we might want to have an "is nested" precondition for pp_token_list
that might have to hold for calls to token_printer::print_tokens vfuncs. We
might want a validator for pp_token_list.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116792
[Bug 116792] RFE: should we be able to generate diagnostics in HTML format?
