Bug #17645: `shorten-64-to-32` errors with Clang 11.1.0 for 32-bit targets w.r.t. `time_t` (maybe Y2038?)

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 3.1.0dev (2021-02-19 master a319c04818) [i686-linux]
• Backport: 2.5: DONTNEED, 2.6: DONTNEED, 2.7: DONE, 3.0: DONE

[xtkoba (Tee KOBAYASHI)]

Targeting to i686-linux-gnu or armv7a-linux-gnu, Clang/LLVM 11.1.0 fails to compile file.c:

compiling file.c
file.c:1256:44: error: implicit conversion loses integer precision: 'const __s64' (aka 'const long long') to 'time_t' (aka 'long') [-Werror,-Wshorten-64-to-32]
    return rb_time_nano_new(stx->stx_btime.tv_sec, stx->stx_btime.tv_nsec);
           ~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~^~~~~~
1 error generated.
make: *** [Makefile:428: file.o] Error 1

A workaround is to explicitly cast the first argument of rb_time_nano_new, as in the attached patch. It will work until 2038.

[xtkoba (Tee KOBAYASHI)]

• Subject changed from `make` stops at `file.c` with Clang 11.1.0 for 32-bit GNU/Linux targets (maybe Y2038?) to `shorten-64-to-32` errors with Clang 11.1.0 for 32-bit targets w.r.t. `time_t` (maybe Y2038?)`shorten-64-to-32` errors with Clang 11.1.0 for 32-bit targets w.r.t. `time_t` (maybe Y2038?)

I encountered another compilation error of shorten-64-to-32 for 32-bit musl targets, where time_t is 64-bit. A workaround:

--- a/thread_sync.c
+++ b/thread_sync.c
@@ -560,7 +560,7 @@ rb_mutex_sleep(VALUE self, VALUE timeout
 
     RUBY_VM_CHECK_INTS_BLOCKING(GET_EC());
     time_t end = time(0) - beg;
-    return INT2FIX(end);
+    return INT2FIX((long)end);
 }
 
 /*

This case, however, does not seem to cause an Y2038 problem in the real world, as the value of end is a time difference in seconds and will almost always be less than 2^31. We need only to suppress the compilation error.

[xtkoba (Tee KOBAYASHI)]

• Subject changed from `shorten-64-to-32` errors with Clang 11.1.0 for 32-bit targets w.r.t. `time_t` (maybe Y2038?)`shorten-64-to-32` errors with Clang 11.1.0 for 32-bit targets w.r.t. `time_t` (maybe Y2038?) to `shorten-64-to-32` errors with Clang 11.1.0 for 32-bit targets w.r.t. `time_t` (maybe Y2038?)

[nobu (Nobuyoshi Nakada)]

Yet another solution would be to define rb_time64_t for rb_time_nano_new (and rb_time_new), but may be an overkill.

[xtkoba (Tee KOBAYASHI)]

• Status changed from Open to Closed

Applied in changeset git|3ac28de5414c81a0da8c8ec969ea74db5a0f1b0f.

---

Explicitly cast __s64 to time_t [Bug #17645]

A workaround of shorten-64-to-32 error where 32-bit linux.

[nobu (Nobuyoshi Nakada)]

• Backport changed from 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN, 3.0: UNKNOWN to 2.5: DONTNEED, 2.6: DONTNEED, 2.7: REQUIRED, 3.0: REQUIRED

[naruse (Yui NARUSE)]

• Backport changed from 2.5: DONTNEED, 2.6: DONTNEED, 2.7: REQUIRED, 3.0: REQUIRED to 2.5: DONTNEED, 2.6: DONTNEED, 2.7: REQUIRED, 3.0: DONE

ruby_3_0 ff2ea4daeb795ebb3d0afa8c57d86a4af9757e50 merged revision(s) 2a6bfd22468343003463e0cbf91953a01b0dbba5,3ac28de5414c81a0da8c8ec969ea74db5a0f1b0f.

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 2.5: DONTNEED, 2.6: DONTNEED, 2.7: REQUIRED, 3.0: DONE to 2.5: DONTNEED, 2.6: DONTNEED, 2.7: DONE, 3.0: DONE

ruby_2_7 9994860dcb56697fad1799128d1493adc06b23c7 merged revision(s) 2a6bfd22468343003463e0cbf91953a01b0dbba5,3ac28de5414c81a0da8c8ec969ea74db5a0f1b0f.