GCC Bugs

The latest version of this document is always available at http://www.gnu.org/software/gcc/bugs.html.

Table of Contents

Reporting Bugs

Our preferred way of receiving bugs is via the GCC GNATS bug reporting system.

Before you report a bug, please check the list of well-known bugs and, if possible in any way, try a current development snapshot. If you want to report a bug with versions of GCC before 3.0 we strongly recommend upgrading to the current release first.

Before reporting that GCC compiles your code incorrectly, please compile it with gcc -Wall and see whether this shows anything wrong with your code that could be the cause instead of a bug in GCC.

Summarized bug reporting instructions

After this summary, you'll find detailed bug reporting instructions, that explain how to obtain some of the information requested in this summary.

What we need

Please include in your bug report all of the following items, the first three of which can be obtained from the output of gcc -v:

What we do not want

Where to post it

Please submit your bug report directly to the GCC GNATS bug database. Only if this is not possible, mail all information to bug-gcc@gnu.org or gcc-bugs@gcc.gnu.org.

The GCC lists have message size limits (200 kbytes) and bug reports over those limits will currently be bounced. If your bug is larger than that, please post it using the GCC GNATS bug database.

Detailed bug reporting instructions

Please refer to the next section when reporting bugs in GNAT, the Ada compiler.

In general, all the information we need can be obtained by collecting the command line below, as well as its output and the preprocessed file it generates.

gcc -v -save-temps all-your-options source-file

Typically the preprocessed file (extension .i for C or .ii for C++) will be large, so please compress the resulting file with one of the popular compression programs such as bzip2, gzip, zip or compress (in decreasing order of preference). Use maximum compression (-9) if available. Please include the compressed preprocessor output in your bug report, even if the source code is freely available elsewhere; it makes the job of our volunteer testers much easier.

The only excuses to not send us the preprocessed sources are (i) if you've found a bug in the preprocessor, or (ii) if you've reduced the testcase to a small file that doesn't include any other file. If you can't post the preprocessed sources because they're proprietary code, then try to create a small file that triggers the same problem.

Since we're supposed to be able to re-create the assembly output (extension .s), you usually should not include it in the bug report, although you may want to post parts of it to point out assembly code you consider to be wrong.

Whether to use MIME attachments or uuencode is up to you. In any case, make sure the compiler command line, version and error output are in plain text, so that we don't have to decode the bug report in order to tell who should take care of it. A meaningful subject indicating language and platform also helps.

Please avoid posting an archive (.tar, .shar or .zip); we generally need just a single file to reproduce the bug (the .i/.ii preprocessed file), and, by storing it in an archive, you're just making our volunteers' jobs harder. Only when your bug report requires multiple source files to be reproduced should you use an archive. In any case, make sure the compiler version, error message, etc, are included in the body of your bug report as plain text, even if needlessly duplicated as part of an archive.

If you fail to supply enough information for a bug report to be reproduced, someone will probably ask you to post additional information (or just ignore your bug report, if they're in a bad day, so try to get it right on the first posting :-). In this case, please post the additional information to the bug reporting mailing list, not just to the person who requested it, unless explicitly told so. If possible, please include in this follow-up all the information you had supplied in the incomplete bug report (including the preprocessor output), so that the new bug report is self-contained.

Detailed bug reporting instructions for GNAT

See the previous section for bug reporting instructions for GCC language implementations other than Ada.

Bug reports have to contain at least the following information in order to be useful:

If your code depends on additional source files (usually package specifications), submit the source code for these compilation units in a single file that is acceptable input to gnatchop, i.e. contains no non-Ada text. If the compilation terminated normally, you can usually obtain a list of dependencies using the "gnatls -d main_unit" command, where main_unit is the file name of the main compilation unit (which is also passed to gcc).

If you report a bug which causes the compiler to print a bug box, include that bug box in your report, and do not forget to send all the source files listed after the bug box along with your report.

If you use gnatprep, be sure to send in preprocessed sources (unless you have to report a bug in gnatprep).

When you have checked that your report meets these criteria, please submit it accoding to our generic instructions. (If you use a mailing list for reporting, please include an "[Ada]" tag in the subject.)

Managing Bugs (GNATS and the test-suite)

This section contains information mostly intended for GCC contributors.

If you find a bug, but you are not fixing it (yet):

  1. Create a (minimal) test-case.
  2. Add the test-case to our test-suite, marking it as XFAIL unless the bug is a regression.
  3. Add a bug report referencing the test-case to GNATS.

If you fix a bug for which there is already a GNATS entry:

  1. Remove the XFAIL on the test-case.
  2. Close the bug report in GNATS.

If you find a bug, and you are fixing it right then:

  1. Create a (minimal) test-case.
  2. Add the test-case to our test-suite, marking it as PASS.
  3. Check in your fixes.

Frequently Reported Bugs in GCC 3.0

General

The following bugs are very frequently reported.

Fortran

Fortran bugs are documented in the G77 manual rather than explicitly listed here. Please see Known Causes of Trouble with GNU Fortran in the G77 manual.

C

The following are not bugs in the C compiler, but are reported often enough to warrant a mention here.

Cannot initialize a static variable with stdin.

This has nothing to do with GCC, but people ask us about it a lot. Code like this:

#include <stdio.h>

FILE *yyin = stdin;

will not compile with GNU libc (GNU/Linux libc6), because stdin is not a constant. This was done deliberately, to make it easier to maintain binary compatibility when the type FILE needs to be changed. It is surprising for people used to traditional Unix C libraries, but it is permitted by the C standard.

This construct commonly occurs in code generated by old versions of lex or yacc. We suggest you try regenerating the parser with a current version of flex or bison, respectively. In your own code, the appropriate fix is to move the initialization to the beginning of main.

There is a common misconception that the GCC developers are responsible for GNU libc. These are in fact two entirely separate projects; please check the GNU libc web pages for details.

Cannot use preprocessor directive in macro arguments.

Let me guess... you wrote code that looks something like this:

  memcpy(dest, src,
#ifdef PLATFORM1
	 12
#else
	 24
#endif
	);

and you got a whole pile of error messages:

test.c:11: warning: preprocessing directive not recognized within macro arg
test.c:11: warning: preprocessing directive not recognized within macro arg
test.c:11: warning: preprocessing directive not recognized within macro arg
test.c: In function `foo':
test.c:6: undefined or invalid # directive
test.c:8: undefined or invalid # directive
test.c:9: parse error before `24'
test.c:10: undefined or invalid # directive
test.c:11: parse error before `#'

Update: As of GCC 3.2 this kind of construct is always accepted and CPP will probably do what you expect, but see the manual for detailed semantics.

However, versions of GCC prior to 3.2 did not allow you to put #ifdef (or any other directive) inside the arguments of a macro. Your C library's <string.h> happens to define memcpy as a macro - this is perfectly legitimate. The code therefore would not compile.

This kind of code is not portable. It is "undefined behavior" according to the C standard; that means different compilers will do different things with it. It is always possible to rewrite code which uses conditionals inside macros so that it doesn't. You could write the above example

#ifdef PLATFORM1
   memcpy(dest, src, 12);
#else
   memcpy(dest, src, 24);
#endif

This is a bit more typing, but I personally think it's better style in addition to being more portable.

In recent versions of glibc, printf is among the functions which are implemented as macros.

C++

This is the list of bugs (and non-bugs) in g++ (aka GNU C++) that are reported very often, but not yet fixed. While it is certainly better to fix bugs instead of documenting them, this document might save people the effort of writing a bug report when the bug is already well-known. How to report bugs tells you how to report a bug.

There are many reasons why reported bugs don't get fixed. It might be difficult to fix, or fixing it might break compatibility. Often, reports get a low priority when there is a simple work-around. In particular, bugs caused by invalid C++ code have a simple work-around, fix the code. Now that there is an agreed ISO/ANSI standard for C++, the compiler has a definitive document to adhere to. Earlier versions might have accepted source code that is no longer C++. This means that code which might have `worked' in a previous version, is now rejected. You should update your code to be C++.

You should try to use the latest stable release of the GNU C++ compiler. This is currently 3.0. Many commonly reported bugs in earlier releases are fixed in that version.

Common problems updating from G++ 2.95 to G++ 3.0

G++ 3.0 conforms much closer to the ISO C++ standard (available at http://www.ncits.org/cplusplus.htm).

We have also implemented some of the core and library defect reports (available at http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/cwg_defects.html & http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/lwg-defects.html respectively).

This means you may get lots of errors about things like strcmp not being found. You've most likely forgotton to tell the compiler to look in the std:: namespace. There are several ways to do this,

ABI bugs

3.0 has a new ABI, which affects class layout, function mangling and calling conventions. We had intended it to be complete, unfortunately some issues have come to light, too late to fix in the 3.0 series. Rather than fix them in subsequent 3.0 dot releases we have decided to fix these in the 3.1 development series. History has shown that it was a mistake to make ABI changes in dot releases. Also, the standard library will also change in the 3.1 series, which is effective as changing the ABI.

Complicated virtual inheritance segfaults
Some instances of complicated virtual inheritance crash when executing the constructor. We believe some cases to be already existing in g++ 2.95, bug that generated incorrect code that didn't happen to crash. Other cases concern polymophic virtual base classes with no data members (so called nearly empty bases). You might be able to work around the problem by adding a dummy data member. GNATS PRs 3598, 3290, 3145 & 2811 track these issues.
Base classes with operator delete [] (void *, size_t) sometimes did not get the correct size passed when deleting a derived object
This is a long standing G++ bug, that should have been fixed by the new ABI. Unfortunately it did not, and requires an ABI change to fix. Also, a related, but more obscure case is when both that and an operator delete[] (void *) is provided. Although the produced code functions, it is incompatible with the ABI. GNATS PR 113 tracked this, which is now fixed in the 3.1 tree with this patch.
Covariant return types
We do not implement non-trivial covariant returns. We also generate incorrect virtual function tables for trivial covariance. Although trivial covariance will work, it is incompatible with the ABI. GNATS PR 3706 tracks this problem.

Non-bugs

Here are some features that have been reported as bugs, but are not.

Nested classes can access private types of the containing class.

G++ now implements type access control on member types. Defect report 45 clarifies that nested classes are members of the class they are nested in, and so are granted access to private members of that class.

Classes in exception specifiers must be complete types.

[15.4]/1 tells you that you cannot have an incomplete type, or pointer to incomplete (other than cv void *) in an exception specification.

G++ emits two copies of constructors and destructors.

In general there are three types of constructors (and destructors).

  1. The complete object constructor/destructor.
  2. The base object constructor/destructor.
  3. The allocating destructor/deallocating destructor.

The first two are different, when virtual base classes are involved. In some cases we can do better, and this is logged in GNATS.

Exceptions don't work in multithreaded applications.

You need to rebuild g++ and libstdc++ with --enable-threads. Remember, c++ exceptions are not like hardware interrupts. You cannot throw an exception in one thread and catch it in another. You cannot throw an exception from a signal handler, and catch it in the main thread.

Global destructors are not run in the correct order.

Global destructors should be run in the reverse order of their constructors completing. In most cases this is the same as the reverse order of constructors starting, but sometimes it is different, and that is important. You need to compile and link your programs with --use-cxa-atexit. We have not turned this switch on by default, as it requires a cxa aware runtime library (libc, glibc, or equivalent).

Problems with floating point computations.

In a number of cases, GCC appears to perform floating point computations incorrectly. For example, the program

#include <iostream>

int main() {

double min = 0.0;
double max = 0.5;
double width = 0.01;
std::cout << (int)(((max - min) / width) - 1) << std::endl;

}

might print 50 on some systems and optimization levels, and 51 on others.

The is the result of rounding: The computer cannot represent all real numbers exactly, so it has to use approximations. When computing with approximation, the computer needs to round to the nearest representable number.

This is not a bug in the compiler, but an inherent limitation of the float and double types. Please study this paper for more information.

Missing features

We know some things are missing from G++.

The export keyword is not implemented.

Most C++ compilers (G++ included) do not yet implement export, which is necessary for separate compilation of template declarations and definitions. Without export, a template definition must be in scope to be used. The obvious workaround is simply to place all definitions in the header itself. Alternatively, the compilation unit containing template definitions may be included from the header.

Two stage lookup in templates is not implemented.

[14.6] specifies how names are looked up inside a template. G++ does not do this correctly, but for most templates this will not be noticeable.

Parse errors for "simple" code

Up to and including GCC 3.0, the compiler will give "parse error" for seemingly simple code, such as 
struct A{
  A();
  A(int);
  void func();
};

struct B{
  B(A);
  B(A,A);
  void func();
};

void foo(){
  B b(A(),A(1));     //Variable b, initialized with two temporaries
  B(A(2)).func();    //B temporary, initialized with A temporary
}
The problem is that GCC starts to parse the declaration of b as a function b returning B, taking a function returning A as an argument. When it sees the 1, it is too late. The work-around in these cases is to add additional parentheses around the expressions that are mistaken as declarations: 
  (B(A(2))).func();
Sometimes, even that is not enough; to show the compiler that this should be really an expression, a comma operator with a dummy argument can be used: 
  B b((0,A()),A(1));

Another example is the parse error for the return statement in

struct A{};

struct B{
  A a;
  A f1(bool);
};

A B::f1(bool b)
{
  if (b)
    return (A()); 
  return a;
}

The problem is that the compiler interprets A() as a function (taking no arguments, returning A), and (A()) as a cast - with a missing expression, hence the parse error. The work-around is to omit the parentheses:

  if (b)
    return A();

This problem occurs in a number of variants; in throw statements, people also frequently put the object in parentheses. The exact error also somewhat varies with the compiler version. The work-arounds proposed do not change the semantics of the program at all; they make them perhaps less readable.

Optimization at -O3 takes a very long time

At -O3, all functions are candidates for inlining. The heuristic used has some deficiencies which show up when allowed such freedom. This is g++ specific, as it has an earlier inliner than gcc.