1 Reporting Bugs in the GNU Compiler Collection for Ubuntu
2 ========================================================
4 Before reporting a bug, please
5 ------------------------------
7 - Check that the behaviour really is a bug. Have a look into some
8 ANSI standards document.
10 - Check the list of well known bugs: http://gcc.gnu.org/bugs.html#known
12 - Try to reproduce the bug with a current GCC development snapshot. You
13 usually can get a recent development snapshot from the gcc-snapshot
14 package in the current development distribution.
16 See: http://archive.ubuntu.com/ubuntu/pool/universe/g/gcc-snapshot/
18 - Try to find out if the bug is a regression (an older GCC version does
21 - Check if the bug is already reported in the bug tracking systems.
23 Ubuntu: https://bugs.launchpad.net/~ubuntu-toolchain/+packagebugs
24 Debian: http://bugs.debian.org/debian-gcc@lists.debian.org
25 Upstream: http://gcc.gnu.org/bugzilla/
31 Please report bugs found in the packaging of GCC to Launchpad. See below
32 how issues should be reported.
34 Ubuntu's current policy is to closely follow the upstream development and
35 only apply a minimal set of patches (which are summarized in the README.Debian
38 If you think you have found an upstream bug, you did check the section
39 above ("Before reporting a bug") and are able to provide a complete bug
40 report (see below "How to report a bug"), then you may help the Ubuntu
41 GCC package maintainers, if you report the bug upstream and then submit
42 a bug report to Launchpad and tell us the upstream report number.
43 This way you are able to follow the upstream bug handling as well. If in
44 doubt, report the bug to Launchpad (but read "How to report a bug" below).
46 Report the issue to https://bugs.launchpad.net/ubuntu/+source/gcc-4.7.
52 There are complete instructions in the gcc info manual (found in the
53 gcc-doc package), section Bugs.
55 The manual can be read using M-x info in Emacs, or if the GNU info
56 program is installed on your system by info --node "(gcc)Bugs". Or see
57 the file BUGS included with the gcc source code.
59 Online bug reporting instructions can be found at
61 http://gcc.gnu.org/bugs.html
63 [Some paragraphs taken from the above URL]
65 The main purpose of a bug report is to enable us to fix the bug. The
66 most important prerequisite for this is that the report must be
67 complete and self-contained, which we explain in detail below.
69 Before you report a bug, please check the list of well-known bugs and,
70 if possible in any way, try a current development snapshot.
72 Summarized bug reporting instructions
73 -------------------------------------
77 Please include in your bug report all of the following items, the
78 first three of which can be obtained from the output of gcc -v:
80 * the exact version of GCC;
82 * the options given when GCC was configured/built;
83 * the complete command line that triggers the bug;
84 * the compiler output (error messages, warnings, etc.); and
85 * the preprocessed file (*.i*) that triggers the bug, generated by
86 adding -save-temps to the complete compilation command, or, in
87 the case of a bug report for the GNAT front end, a complete set
88 of source files (see below).
92 * A source file that #includes header files that are left out
93 of the bug report (see above)
94 * That source file and a collection of header files.
95 * An attached archive (tar, zip, shar, whatever) containing all
96 (or some :-) of the above.
97 * A code snippet that won't cause the compiler to produce the
98 exact output mentioned in the bug report (e.g., a snippet with
99 just a few lines around the one that apparently triggers the
100 bug, with some pieces replaced with ellipses or comments for
101 extra obfuscation :-)
102 * The location (URL) of the package that failed to build (we won't
103 download it, anyway, since you've already given us what we need
104 to duplicate the bug, haven't you? :-)
105 * An error that occurs only some of the times a certain file is
106 compiled, such that retrying a sufficient number of times
107 results in a successful compilation; this is a symptom of a
108 hardware problem, not of a compiler bug (sorry)
109 * E-mail messages that complement previous, incomplete bug
110 reports. Post a new, self-contained, full bug report instead, if
111 possible as a follow-up to the original bug report
112 * Assembly files (*.s) produced by the compiler, or any binary files,
113 such as object files, executables, core files, or precompiled
115 * Duplicate bug reports, or reports of bugs already fixed in the
116 development tree, especially those that have already been
117 reported as fixed last week :-)
118 * Bugs in the assembler, the linker or the C library. These are
119 separate projects, with separate mailing lists and different bug
121 * Bugs in releases or snapshots of GCC not issued by the GNU
122 Project. Report them to whoever provided you with the release
123 * Questions about the correctness or the expected behavior of
124 certain constructs that are not GCC extensions. Ask them in
125 forums dedicated to the discussion of the programming language
128 Known Bugs and Non-Bugs
129 -----------------------
131 [Please see /usr/share/doc/gcc/FAQ or http://gcc.gnu.org/faq.html first]
134 C++ exceptions don't work with C libraries
135 ------------------------------------------
137 [Taken from the closed bug report #22769] C++ exceptions don't work
138 with C libraries, if the C code wasn't designed to be thrown through.
139 A solution could be to translate all C libraries with -fexceptions.
140 Mostly trying to throw an exception in a callback function (qsort,
141 Tcl command callbacks, etc ...). Example:
149 int SortCondition(void const*, void const*)
151 printf("throwing 'sortcondition' exception\n");
155 int main(int argc, char *argv[])
160 SortCondition(NULL,NULL);
162 printf("caught test-sortcondition exception\n");
165 qsort(&list, sizeof(list)/sizeof(list[0]),sizeof(list[0]),
168 printf("caught real-sortcondition exception\n");
173 Andrew Macleod <amacleod@cygnus.com> responded:
175 When compiled with the table driven exception handling, exception can only
176 be thrown through functions which have been compiled with the table driven EH.
177 If a function isn't compiled that way, then we do not have the frame
178 unwinding information required to restore the registers when unwinding.
180 I believe the setjmp/longjmp mechanism will throw through things like this,
181 but its produces much messier code. (-fsjlj-exceptions)
183 The C compiler does support exceptions, you just have to turn them on
186 Your main options are to:
187 a) Don't use callbacks, or at least don't throw through them.
188 b) Get the source and compile the library with -fexceptions (You have to
189 explicitly turn on exceptions in the C compiler)
190 c) always use -fsjlj-exceptions (boo, bad choice :-)
193 g++: "undefined reference" to static const array in class
194 ---------------------------------------------------------
196 The following code compiles under GNU C++ 2.7.2 with correct results,
197 but produces the same linker error with GNU C++ 2.95.2.
198 Alexandre Oliva <oliva@lsd.ic.unicamp.br> responded:
200 All of them are correct. A static data member *must* be defined
201 outside the class body even if it is initialized within the class
202 body, but no diagnostic is required if the definition is missing. It
203 turns out that some releases do emit references to the missing symbol,
204 while others optimize it away.
213 static const unsigned char Jam_signature[4] = "JAM";
216 Test::Test(const char *q)
218 if (memcmp(q, Jam_signature, sizeof(Jam_signature)) != 0)
219 cerr << "Hello world!\n";
228 g++: g++ causes passing non const ptr to ptr to a func with const arg
229 to cause an error (not a bug)
230 ---------------------------------------------------------------------
235 void test(const char **b){
244 g++ const.cc -o const
245 const.cc: In function int main():
246 const.cc:7: passing char ** as argument 1 of test(const char **) adds cv-quals without intervening const
247 make: *** [const] Error 1
249 Answer from "Martin v. Loewis" <martin@loewis.home.cs.tu-berlin.de>:
251 > ok... maybe I missed something.. I haven't really kept up with the latest in
252 > C++ news. But I've never heard anything even remotly close to passing a non
253 > const var into a const arg being an error before.
255 Thanks for your bug report. This is a not a bug in the compiler, but
256 in your code. The standard, in 4.4/4, puts it that way
258 # A conversion can add cv-qualifiers at levels other than the first in
259 # multi-level pointers, subject to the following rules:
260 # Two pointer types T1 and T2 are similar if there exists a type T and
261 # integer n > 0 such that:
262 # T1 is cv(1,0) pointer to cv(1,1) pointer to ... cv(1,n-1)
263 # pointer to cv(1,n) T
265 # T2 is cv(2,0) pointer to cv(2,1) pointer to ... cv(2,n-1)
266 # pointer to cv(2,n) T
267 # where each cv(i,j) is const, volatile, const volatile, or
268 # nothing. The n-tuple of cv-qualifiers after the first in a pointer
269 # type, e.g., cv(1,1) , cv(1,2) , ... , cv(1,n) in the pointer type
270 # T1, is called the cv-qualification signature of the pointer type. An
271 # expression of type T1 can be converted to type T2 if and only if the
272 # following conditions are satisfied:
273 # - the pointer types are similar.
274 # - for every j > 0, if const is in cv(1,j) then const is in cv(2,j) ,
275 # and similarly for volatile.
276 # - if the cv(1,j) and cv(2,j) are different, then const is in every
277 # cv(2,k) for 0 < k < j.
279 It is the last rule that your code violates. The standard gives then
280 the following example as a rationale:
282 # [Note: if a program could assign a pointer of type T** to a pointer
283 # of type const T** (that is, if line //1 below was allowed), a
284 # program could inadvertently modify a const object (as it is done on
285 # line //2). For example,
287 # const char c = 'c';
289 # const char** pcc = &pc; //1: not allowed
291 # *pc = 'C'; //2: modifies a const object
295 If you question this line of reasoning, please discuss it in one of
296 the public C++ fora first, eg. comp.lang.c++.moderated, or
300 cpp removes blank lines
301 -----------------------
303 With the new cpp, you need to add -traditional to the "cpp -P" args, else
304 blank lines get removed.
306 [EDIT ME: scan Debian bug reports and write some nice summaries ...]