Need help with providing a sensible bug report.

Colin Close itchka@compuserve.com
Fri Dec 20 00:50:00 GMT 2019


Hi, I have found a regression in gcc and have bisected it to a particular commit.
Having read the bug reporting page there's nothing that really covers my situation.
I can can tell you that commit xxx breaks some gecko source code giving an internal compiler error and a line number in an include file.
If I generate pre-processed source using g++ -E.  I can successfully compile this into an object file without the error showing itself.
The error though is reproducible when compiling in the normal manner in the source tree and can be invoked by the compilation of a single file.
When building the source tree the error can occur in a more random manner in that it does not occur on the same source file every time but the source file that fails always contains the same #include and reports the same error and line number every time.

The faulty commit is 067e9a502dcaabcecda20a00b3f7c6160d8eea93
The error message is.
Development/bluegriffon/BUILD/gecko-dev-042b84a/opt/dist/include/jsfriendapi.h:2005:1: internal compiler error: in write_type, at cp/mangle.c:2073
2005 | JS_GetArrayBufferData(JSObject* obj, bool* isSharedMemory, const JS::AutoCheckCannotGC&);
The file is attached but I am unable to attach the pre-processed source as it exceeds the 400k mail size allowance.

Before filing a bug it would be helpful to know if I have provided sufficient information or whether there is more that I should do to aid those may fix the problem.

Colin Close
-------------- next part --------------
A non-text attachment was scrubbed...
Name: jsfriendapi.h
Type: text/x-chdr
Size: 103832 bytes
Desc: not available
URL: <https://gcc.gnu.org/pipermail/gcc-help/attachments/20191220/3bea682d/attachment.bin>


More information about the Gcc-help mailing list