Hi, I found a bug in `zic.c` and would like to propose a fix. Please see the attached file.
Best regards, Dmitry Khalanskiy.
From 622d11fdb9dfc0b02e3db1a68a12020018dc4778 Mon Sep 17 00:00:00 2001 From: Dmitry Khalanskiy <[email protected]> Date: Mon, 2 Nov 2026 10:48:06 +0100 Subject: [PATCH] Fix reading uninitialized data When the `-r` specifies a high bound, `outzone` omits a rule string and instead emits a fake transition 402 after the last actually observed one. If there were no recorded transitions, the fake transition may or may not get created, as this depends on the state of memory that does not get cleaned after processing earlier time zones. Specifically, `lastat` gets set to `attypes`, which is a non-`null` pointer, so the check `!lastat` always fails, leading to the read of `lastat->at` and executing logic based on the values within. To reproduce the uninitialized read, run with the arguments `-r /@12717475200 tzdata.zi` and read the compiled `Etc/UTC`. It should include the fake transition, since `2373-01-01` is exactly 402 years after the last recorded transition (`1970-01-01`), but it does not. --- zic.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/zic.c b/zic.c index cd0dcac2..a93be126 100644 --- a/zic.c +++ b/zic.c @@ -3733,7 +3733,7 @@ outzone(const struct zone *zpfirst, ptrdiff_t zonecount) for (lastat = attypes, i = 1; i < timecnt; i++) if (attypes[i].at > lastat->at) lastat = &attypes[i]; - if (!lastat || lastat->at < rpytime(&xr, max_year - 1)) { + if (0 == timecnt || lastat->at < rpytime(&xr, max_year - 1)) { addtt(rpytime(&xr, max_year + 1), lastat ? lastat->type : defaulttype); attypes[timecnt - 1].dontmerge = true; -- 2.54.0
