[llvm-dev] [cfe-dev] Location (in Source Code) of String Literal in MacroExpansion

chiasa.men via llvm-dev llvm-dev at lists.llvm.org
Thu Aug 26 23:27:29 PDT 2021


Just a short feedback on that. I played a bit with you code and got the needed 
information seems to be 

void MacroExpands(const Token &MacroName, const MacroDefinition &MD, 
SourceRange Range, const MacroArgs *Args) override {
    for (unsigned i=0;i<Args->getNumMacroArguments ();++i){
        Args->getUnexpArgument(i)->getLocation().printToString(SM)<<":"<<Args-
>getUnexpArgument(i)->getEndLoc().printToString(SM)<<"\n";
    }
}

I went ahead and basically copied and pasted your code in order to add a new 
collection for unexpanded arguments.

In contrast to your code the tokenWatcher does not make sense in this regard 
since it will probably lex the preprocessed hence expanded source code.

Thanks for your work

On Donnerstag, 26. August 2021 17:00:09 CEST chiasa.men via llvm-dev wrote:
> Thanks you were right, that fixed the minor printing issue.
> But the actual problem of string literals in macros still persists. Maybe
> someone else has an idea on that. The information isn't in reach of your
> MacroExpansionContext/MacroExpansionRangeRecorder, is it?
> 
> On Donnerstag, 26. August 2021 12:32:22 CEST benicsbalazs at gmail.com wrote:4
> > Oh, now I see why you were puzzled!
> > 
> > 
> > 
> > `printf()` expects a null-terminated string (“%s”), but you pass an
> > llvm::StringRef instead, which is not null terminated!
> > 
> > Thus, you will end up with undefined behavior, which ends up printing till
> > the end of the file.
> > 
> > 
> > 
> > I hope this is it!
> > 
> > 
> > 
> > Balazs
> > 
> > 
> > 
> > From: benicsbalazs at gmail.com <benicsbalazs at gmail.com>
> > Sent: 2021. augusztus 26., csütörtök 12:25
> > To: 'chiasa.men' <chiasa.men at web.de>
> > Cc: llvm-dev at lists.llvm.org
> > Subject: RE: [cfe-dev] Location (in Source Code) of String Literal in
> > MacroExpansion
> > 
> > 
> > 
> > Hi Chiasa,
> > 
> > 
> > 
> > For the example you attached the MacroExpansionContext seems to work as I
> > intended.
> > 
> > It records what range gets replaced by what tokens at the end of the
> > preprocessing, thus we get
> > 
> > `CONCATMACRO("TEST")` -> ` L"TEST"`
> > 
> > 
> > 
> > About `Lexer::getSourceText()` I’m not sure. I haven’t used that. The
> > `at()` function was designed as a helper defining the unit tests. I might
> > misused something when I created this code & tests, let me know!
> > 
> > 
> > 
> > Balazs
> > 
> > 
> > 
> > From: cfe-dev <cfe-dev-bounces at lists.llvm.org
> > <mailto:cfe-dev-bounces at lists.llvm.org> > On Behalf Of chiasa.men via
> > cfe-dev Sent: 2021. augusztus 26., csütörtök 10:03
> > To: cfe-dev <cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org> >
> > Cc: llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
> > Subject: [cfe-dev] Location (in Source Code) of String Literal in
> > MacroExpansion
> > 
> > 
> > 
> > https://gist.github.com/T35R6braPwgDJKq/6439fda090e4ee8440d726f8dafa4dbb#f
> > il e-mectest-cpp
> > 
> > 
> > 
> > Above I modified the MacroExpansionContext unittest.
> > 
> > The relevant code:
> > 
> > //--------------
> > 
> > TEST_F(MacroExpansionContextTest, Custom) {
> > 
> >   // From the GCC website.
> >   
> >   const auto Ctx = getMacroExpansionContextFor(R"code(
> > 
> > #define CONCATMACRO(parm) L##parm
> > 
> > int f(wchar_t *b){}
> > 
> > int main()
> > 
> > {
> > 
> >     int a= f(CONCATMACRO("TEST"));
> >     
> >     int b= f(L"TEST");
> >     
> >     return 0;
> > 
> > }
> > 
> > )code");
> > 
> >     Ctx->dumpExpansionRanges();
> >     
> >     //> :6:14, :6:33
> >     
> >     Ctx->dumpExpandedTexts();
> >     
> >     //> :6:14 -> 'L"TEST"'
> >     
> >     printf("Exp: %s\n",Ctx->getExpandedText(at(6, 14)).getValue());
> >     
> >     //Exp: L"TEST"
> >     
> >     printf("Org: %s\n",Ctx->getOriginalText(at(6, 14)).getValue());
> >     
> >     //Org: CONCATMACRO("TEST"));
> >     
> >     //    int b= f(L"TEST");
> >     
> >     //    return 0;
> >     
> >     //}
> >     
> >     StringRef sourceText =
> > 
> > clang::Lexer::getSourceText(CharSourceRange(SourceRange(at(6, 14),at(6,
> > 33)),false),SourceMgr, LangOpts);
> > 
> >     printf("sourceText: %s\n",sourceText);
> >     
> >     // sourceText: CONCATMACRO("TEST"));
> >     
> >     //     int b= f(L"TEST");
> >     
> >     //     return 0;
> >     
> >     // }
> > 
> > }
> > 
> > //--------------
> > 
> > I am interested in getting the range for "TEST" in the SourceText. Since
> > the CONCATMACRO is preprocessed to L"TEST" I thought the
> > MacroExpansionContext would be of help.
> > 
> > It tells about the ExpansionRanges: 6:14, :6:33 which is
> > CONCATMACRO("TEST")
> > 
> > What I would need is: 6:26, 6:32
> > 
> > Obviously I could just calculate a bit like:
> > 
> > range.begin (14) + macronameLen (11) +1 (opening parenthesis) = 26
> > 
> > to
> > 
> > range.end (33) - 1 (closing parenthesis) = 32
> > 
> > 
> > 
> > But would that be reliable and is there no other way?
> > 
> > 
> > 
> > Furthermore: Why do getSourceText/getOriginalText print the whole rest of
> > the source code and not just from (6,14) to (6,33)?


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20210827/35111c15/attachment.html>


More information about the llvm-dev mailing list