Package: cfortran Followup-For: Bug #1114078 User: [email protected] Usertags: patch
Dear Maintainers, In Ubuntu stonking (and Debian sid with GCC 15), cfortran fails to build from source due to C23 prototype semantics and a heap buffer overflow in tests under _FORTIFY_SOURCE=3: 1. GCC 15 defaults to C23 where empty parameter lists in prototypes mean strictly 0 arguments. The internal testsuite in eg/ exercises legacy Fortran/C prototypes and function pointer casts into qsort (__compar_fn_t), resulting in: e2/e2.c:10:28: error: too many arguments to function 'easy_'; expected 0, have 2 q/q.c:15: error: passing 'int (*)(void)' to parameter of type '__compar_fn_t' Setting 'export DEB_CFLAGS_MAINT_APPEND = -std=gnu17' in debian/rules restores GNU17 semantics for building and running the testsuite. 2. In eg/fstr/fstr.c, Pstr allocates malloc(ls > lsave ? ls : lsave) without space for the terminating NUL byte. When strcpy writes lsave+1 bytes, it triggers buffer overflow termination under _FORTIFY_SOURCE=3. Allocating + 1 byte fixes the crash. (Also submitted upstream as PR #3: https://github.com/bastien-roucaries/cfortran/pull/3). Attached below is the patch against cfortran 20210827-1.1. With this patch, all 41 testsuites pass and the package builds cleanly under GCC 15. Thank you, Viktor Pashaiev <[email protected]> --- a/debian/rules +++ b/debian/rules @@ -4,6 +4,7 @@ # Uncomment this to turn on verbose mode. export DH_VERBOSE=1 +export DEB_CFLAGS_MAINT_APPEND = -std=gnu17 %: dh $@ --with autoreconf --- a/eg/fstr/fstr.c +++ b/eg/fstr/fstr.c @@ -16,7 +16,7 @@ if (!s || !save) { save=s; return; } ls = strlen(s ); lsave = strlen(save); -temp = (char *)malloc(ls>lsave?ls:lsave); +temp = (char *)malloc((ls>lsave?ls:lsave) + 1); /* Switch contents of argument with contents of saved string. */ strcpy(temp,save); strcpy(save,s );
