[ https://issues.apache.org/jira/browse/STDCXX-989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12612744#action_12612744 ]
Martin Sebor commented on STDCXX-989: ------------------------------------- Yowza! Looks like we're a few bytes off. On Solaris/SPARC, {{sizeof(jmp_buf)}} is {{12 * sizeof(long)}} bytes, or 48 bytes in ILP32 and 96 in LP64. On Linux, it looks like it's 156 bytes in ILP32 and 200 in LP64. I don't know where the 8 came from... We should probably have a config test for it. > [EDG C++ 3.9] definition of jmp_buf causes buffer overflow > ---------------------------------------------------------- > > Key: STDCXX-989 > URL: https://issues.apache.org/jira/browse/STDCXX-989 > Project: C++ Standard Library > Issue Type: Bug > Components: 18. Language Support > Affects Versions: 4.2.x > Environment: Linux/EDG C++ 3.9 > Reporter: Travis Vitek > Fix For: 4.2.x > > Original Estimate: 1h > Remaining Estimate: 1h > > The {{jmp_buf}} definition provided in {{include/ansi/\_csetjmp.h}} is not as > large as {{setjmp()}} is expecting. This causes a buffer overflow when > running tests that use the test driver ({{rw_test}}), which results in data > corruption. If you run the tests with {{--trace}}, you will see that the > _clause_ field of the output is always corrupt after the first use. > The following shows the expected size for us... > {noformat} > $ cat u.cpp; eccp u.cpp && ./a.out > #include <setjmp.h> > #include <stdio.h> > int main () > { > printf ("%u\n", sizeof (jmp_buf)); > return 0; > } > 156 > {noformat} > The definition of {{jmp_buf}} in {{include/rw/_csetjmp.h}} is a buffer of 8 > bytes. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.