gcc 已知的问题及规避方案 https://gcc.gnu.org/bugs/#known

潘慈

2023-12-01

https://gcc.gnu.org/bugs/#known

GCC Bugs

Reporting Bugs

The main purpose of a bug report is to enable us to fix the bug. The most important prerequisite for this is that the report must be complete and self-contained.

Before you report a bug, please check the list of well-known bugs and, if possible, try a current release or development snapshot.

Before reporting that GCC compiles your code incorrectly, compile it with gcc -Wall -Wextra and see whether this shows anything wrong with your code. Similarly, if compiling with -fno-strict-aliasing -fwrapv -fno-aggressive-loop-optimizations makes a difference, your code probably is not correct.

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:

the exact version of GCC;
the system type;
the options given when GCC was configured/built;
the complete command line that triggers the bug;
the compiler output (error messages, warnings, etc.); and
the preprocessed file (*.i*) that triggers the bug, generated by adding -save-temps to the complete compilation command, or, in the case of a bug report for the GNAT front end, a complete set of source files (see below).

What we do not want

A source file that #includes header files that are left out of the bug report (see above)
That source file and a collection of header files.
An attached archive (tar, zip, shar, whatever) containing all (or some :-) of the above.
A code snippet that won't cause the compiler to produce the exact output mentioned in the bug report (e.g., a snippet with just a few lines around the one that apparently triggers the bug, with some pieces replaced with ellipses or comments for extra obfuscation :-)
The location (URL) of the package that failed to build (we won't download it, anyway, since you've already given us what we need to duplicate the bug, haven't you? :-)
An error that occurs only some of the times a certain file is compiled, such that retrying a sufficient number of times results in a successful compilation; this is a symptom of a hardware problem, not of a compiler bug (sorry)
Assembly files (*.s) produced by the compiler, or any binary files, such as object files, executables, core files, or precompiled header files
Duplicate bug reports, or reports of bugs already fixed in the development tree, especially those that have already been reported as fixed last week :-)
Bugs in the assembler, the linker or the C library. These are separate projects, with separate mailing lists and different bug reporting procedures
Bugs in releases or snapshots of GCC not issued by the GNU Project. Report them to whoever provided you with the release
Questions about the correctness or the expected behavior of certain constructs that are not GCC extensions. Ask them in forums dedicated to the discussion of the programming language

Where to post it

Please submit your bug report directly to the GCC bug tracker.

The GCC bug tracker requires you to create an account with a valid e-mail address. This is not merely to be annoying. It's because in the past spammers have filed fake bug reports, and fake attachments to real bug reports, to distribute malware and to add links to their spam web sites. Requiring a valid e-mail address is a partial deterrent to this. We apologize for the inconvenience.

Detailed bug reporting instructions

Please refer to the next section when reporting bugs in GNAT, the Ada compiler, or to the one after that when reporting bugs that appear when using a precompiled header.

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

The preprocessed source is the basic requirement to fix a bug. However, providing a minimal testcase increases the chances of getting your bug fixed. The only excuses to not send us the preprocessed sources are (i) if you've found a bug in the preprocessor, (ii) if you've reduced the testcase to a small file that doesn't include any other file or (iii) if the bug appears only when using precompiled headers. 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.

Please avoid posting an archive (.tar, .shar or .zip); we generally need just a single file to reproduce the bug (the .i/.ii/.f 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. This is, for example, the case if you are using INCLUDE directives in Fortran code, which are not processed by the preprocessor, but the compiler. In that case, we need the main file and all INCLUDEd files. 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.

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:

the exact version of GCC, as shown by "gcc -v";
the system type;
the options when GCC was configured/built;
the exact command line passed to the gcc program triggering the bug (not just the flags passed to gnatmake, but gnatmake prints the parameters it passed to gcc)
a collection of source files for reproducing the bug, preferably a minimal set (see below);
a description of the expected behavior;
a description of actual behavior.

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 according to our generic instructions. (If you use a mailing list for reporting, please include an "[Ada]" tag in the subject.)

Detailed bug reporting instructions when using a precompiled header

If you're encountering a bug when using a precompiled header, the first thing to do is to delete the precompiled header, and try running the same GCC command again. If the bug happens again, the bug doesn't really involve precompiled headers, please report it without using them by following the instructions above.

If you've found a bug while building a precompiled header (for instance, the compiler crashes), follow the usual instructions above.

If you've found a real precompiled header bug, what we'll need to reproduce it is the sources to build the precompiled header (as a single .i file), the source file that uses the precompiled header, any other headers that source file includes, and the command lines that you used to build the precompiled header and to use it.

Please don't send us the actual precompiled header. It is likely to be very large and we can't use it to reproduce the problem.

Frequently Reported Bugs

There are many reasons why a reported bug doesn'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 code have a simple work-around: fix the code.

G77 bugs were documented under Known Causes of Trouble with GNU Fortran in the G77 manual.

Non-bugs

The following are not actually bugs, but are reported often enough to warrant a mention here.

It is not always a bug in the compiler, if code which "worked" in a previous version, is now rejected. Earlier versions of GCC sometimes were less picky about standard conformance and accepted invalid source code. In addition, programming languages themselves change, rendering code invalid that used to be conforming (this holds especially for C++). In either case, you should update your code to match recent language standards.

General

Problems with floating point numbers - the most often reported non-bug.

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

#include <iostream>

int main()
{
  double a = 0.5;
  double b = 0.01;
  std::cout << (int)(a / b) << std::endl;
  return 0;
}

might print 50 on some systems and optimization levels, and 49 on others.

This 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 floating point types. Please study this paper for more information.

C

Increment/decrement operator ( ++/ --) not working as expected - a problem with many variations.

The following expressions have unpredictable results:

x[i]=++i
foo(i,++i)
i*(++i)                 /* special case with foo=="operator*" */
std::cout << i << ++i   /* foo(foo(std::cout,i),++i)          */

since the i without increment can be evaluated before or after ++i.

The C and C++ standards have the notion of "sequence points". Everything that happens between two sequence points happens in an unspecified order, but it has to happen after the first and before the second sequence point. The end of a statement and a function call are examples for sequence points, whereas assignments and the comma between function arguments are not.

Modifying a value twice between two sequence points as shown in the following examples is even worse:

i=++i
foo(++i,++i)
(++i)*(++i)               /* special case with foo=="operator*" */
std::cout << ++i << ++i   /* foo(foo(std::cout,++i),++i)        */

This leads to undefined behavior (i.e. the compiler can do anything).

Casting does not work as expected when optimization is turned on.

This is often caused by a violation of aliasing rules, which are part of the ISO C standard. These rules say that a program is invalid if you try to access a variable through a pointer of an incompatible type. This is happening in the following example where a short is accessed through a pointer to integer (the code assumes 16-bit shorts and 32-bit ints):

#include <stdio.h>

int main()
{
  short a[2];

  a[0]=0x1111;
  a[1]=0x1111;

  *(int *)a = 0x22222222; /* violation of aliasing rules */

  printf("%x %x\n", a[0], a[1]);
  return 0;
}

The aliasing rules were designed to allow compilers more aggressive optimization. Basically, a compiler can assume that all changes to variables happen through pointers or references to variables of a type compatible to the accessed variable. Dereferencing a pointer that violates the aliasing rules results in undefined behavior.

In the case above, the compiler may assume that no access through an integer pointer can change the array a, consisting of shorts. Thus, printf may be called with the original values of a[0] and a[1]. What really happens is up to the compiler and may change with architecture and optimization level.

Recent versions of GCC turn on the option -fstrict-aliasing (which allows alias-based optimizations) by default with -O2. And some architectures then really print "1111 1111" as result. Without optimization the executable will generate the "expected" output "2222 2222".

To disable optimizations based on alias-analysis for faulty legacy code, the option -fno-strict-aliasing can be used as a work-around.

The option -Wstrict-aliasing (which is included in -Wall) warns about some - but not all - cases of violation of aliasing rules when -fstrict-aliasingis active.

To fix the code above, you can use a union instead of a cast (note that this is a GCC extension which might not work with other compilers):

#include <stdio.h>

int main()
{
  union
  {
    short a[2];
    int i;
  } u;

  u.a[0]=0x1111;
  u.a[1]=0x1111;

  u.i = 0x22222222;

  printf("%x %x\n", u.a[0], u.a[1]);
  return 0;
}

Now the result will always be "2222 2222".

For some more insight into the subject, please have a look at this article.

这个问题导致当前一些DSP芯片中自定义的数据类型，使用指针做强制类型转换时存在运行隐患；

Loops do not terminate

This is often caused by out-of-bound array accesses or by signed integer overflow which both result in undefined behavior according to the ISO C standard. For example

int
SATD (int* diff, int use_hadamard)
{
  int k, satd = 0, m[16], dd, d[16];
  ...
    for (dd=d[k=0]; k<16; dd=d[++k])
      satd += (dd < 0 ? -dd : dd);

accesses d[16] before the loop is exited with the k<16 check. This causes the compiler to optimize away the exit test because the new value of kmust be in the range [0, 15] according to ISO C.

GCC starting with version 4.8 has a new option -fno-aggressive-loop-optimizations that may help here. If it does, then this is a clear sign that your code is not conforming to ISO C and it is not a GCC bug.

Cannot use preprocessor directive in macro arguments.

Let me guess... you used an older version of GCC to compile 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

This is because your C library's <string.h> happens to define memcpy as a macro - which is perfectly legitimate. In recent versions of glibc, for example, printf is among those functions which are implemented as macros.

Versions of GCC prior to 3.3 did not allow you to put #ifdef (or any other preprocessor directive) inside the arguments of a macro. The code therefore would not compile.

As of GCC 3.3 this kind of construct is always accepted and the preprocessor will probably do what you expect, but see the manual for detailed semantics.

However, this kind of code is not portable. It is "undefined behavior" according to the C standard; that means different compilers may 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.

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, 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.

C++

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.