| Ticket UUID: | 176de86b3bb11df158b0245f8706cf9e543ccc19 | ||
| Title: | TEBCResume(): pointer overflow for INST_LREPLACE4 | ||
| Type: | Bug | Created on: | 2023-05-08 16:16:12 |
| Submitter: | chrstphrchvz | Assigned to: | nobody |
| Subsystem: | 17. Commands I-L | Severity: | Minor |
| Priority: | 5 Medium | Last modified: | 2026-11-08 06:44:29 |
| Status: | Closed | Closed by: | chrstphrchvz |
| Resolution: | Duplicate | Closed on: | 2026-11-08 06:44:29 |
| Version: | core-9-0-branch | ||
| Description: | ||||
|
As of [7a2fc29ce038], numNewElems is declared as size_t, which can lead to pointer overflow in OBJ_AT_DEPTH(numNewElems - 1). Example -fsanitize=pointer-overflow error during autoMkindex-3.4: generic/tclExecute.c:5109:9: runtime error: subtraction of unsigned offset from 0x611000099078 overflowed to 0x611000099080 Declaring numToDelete as size_t seems fine, though. | ||||
| User Comments: | ||||
chrstphrchvz added on 2023-09-15 13:28:28:
As of [ba731752d345] I wonder if this issue also affects 8.7 (particularly on 32-bit); so far I have not observed that it does. However fix for this issue may still be applicable to both 8.7 and 9.0. griffin added on 2023-09-16 17:25:35:
Fixed by broader updates to the uses of size_t everywhere. chrstphrchvz added on 2023-09-16 22:10:27:
Can you please elaborate how this was fixed? I still observe this on trunk, since TCL_HASH_TYPE is size_t. griffin added on 2023-09-17 16:22:19:
I cannot reproduce the error. My first attempt produced no error. After review I realized I didn't run the build & test correctly. It this point I cannot get tcl to build correctly with sanitize at all, let alone run. Re-open the bug for someone else to pursue. chrstphrchvz added on 2025-09-19 21:59:01:
This error is still present (rephrased slightly):
and:
I no longer see this error during autoMkindex tests, but there are several other tests which do trigger it; one affected command is chrstphrchvz added on 2026-11-08 06:44:29:
Closing as duplicate in favor of [94c6511e]. | ||||
