diff mbox series

[13/59] Remove alignment assumption from __tzfile_read

Message ID	20250105055750.1668721-14-eggert@cs.ucla.edu (mailing list archive)
State	Failed CI
Headers	DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org A7B093858C56 From: Paul Eggert <eggert@cs.ucla.edu> To: libc-alpha@sourceware.org Cc: Paul Eggert <eggert@cs.ucla.edu> Subject: [PATCH 13/59] Remove alignment assumption from __tzfile_read Date: Sat, 4 Jan 2025 21:56:48 -0800 Message-ID: <20250105055750.1668721-14-eggert@cs.ucla.edu> In-Reply-To: <01207110-bd60-46ae-9c08-fb39c2011067@cs.ucla.edu> References: <01207110-bd60-46ae-9c08-fb39c2011067@cs.ucla.edu> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Precedence: list Errors-To: libc-alpha-bounces~patchwork=sourceware.org@sourceware.org
Series	time: sync mktime from Gnulib \| [00/59] time: sync mktime from Gnulib [01/59] Split intprops.h into two [02/59] Support stdckdint.h internally if older GCC [03/59] Fix mishandling of default DST rule [04/59] Don't worry about !TZDEFAULT [05/59] Support static_assert in pre-C23 C [06/59] Sync mktime.c from Gnulib 2024-10-04 [07/59] Document mktime out-of-range + tm_isdst [08/59] Push tzset lock into callers of time functions [09/59] Omit parse_offset unnecessary tz == '\0' [10/59] Don’t assume TZif bloat is likely [11/59] tzset no longer calls ftello [12/59] Check TZif files more carefully [13/59] Remove alignment assumption from __tzfile_read [14/59] Treat oddball UTC offset strings as zero [15/59] Simplify __tz_compute [16/59] Improve performance in 2 BC [17/59] Don’t mishandle years < 1970 in compute_change [18/59] Match RFC 9636 names [19/59] Trivial tzset_internal speedup [20/59] Fix obscure bug when stdoffset == dstoffset [21/59] Document assumption that NO_DST == 0 [22/59] __use_tzfile is boolean [23/59] Minor tzfile.c clarity improvements [24/59] Stop using zic's -y option [25/59] Fix mishandling of default DST rule [26/59] Simplify treatment of missing and empty TZ [27/59] TZ="" always means UTC sans leap seconds [28/59] Prefer "UTC" for UTC [29/59] Simplify setting tz_rules to UTC [30/59] Update tzname etc. even if TZ is unchanged [31/59] "POSIX TZ" -> "proleptic TZ" [32/59] Reject invalid TZ strings in TZif files [33/59] Reject TZif files containing '\0' in TZ string [34/59] mktime should not consult 'daylight' [35/59] Remove __use_tzfile [36/59] Remove arbitrary limit on TZ, TZDIR lengths [37/59] Tighten setuid TZif file name check [38/59] Avoid a malloc in __tzfile_read [39/59] Free earlier in __tzfile_read [40/59] Improve __tzfile_read TZ="" comment [41/59] Refactor enum tz_rule_type [42/59] __offtime now returns struct tm [43/59] Fix overflow checking in offtime [44/59] Handle TZif time type 0 correctly [45/59] Remove invalid tzset tests from tst-bz29951.c [46/59] tzset, not localtime, now parses TZif TZ spec [47/59] Minor compute_change widening tweak [48/59] Minor SECSPERDAY widening tweak [49/59] Remove leap-second goto from __tzfile_compute [50/59] Remove use_last label from __tzfile_compute [51/59] Refactor ‘if’ in __tzfile_compute [52/59] Refactor localtime ‘if’ in __tzfile_compute [53/59] Refactor post-transition-table ‘if’ [54/59] Refactor __tzfile_compute to avoid two gotos [55/59] Refactor __tzfile_compute index local [56/59] mktime: prefer bool to int [57/59] mktime: port __mktime64 locking to Gnulib [58/59] mktime: improve tm_isdst heuristic [59/59] time: test DST adjustment for DST-less zones

Checks

Context	Check	Description
redhat-pt-bot/TryBot-apply_patch	success	Patch applied to master at the time it was sent

Commit Message

Paul Eggert Jan. 5, 2025, 5:56 a.m. UTC

  * time/tzfile.c (__tzfile_read): Allocate more portably,
removing the need for a couple of _Static_asserts.
---
 time/tzfile.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff mbox series

Patch

diff --git a/time/tzfile.c b/time/tzfile.c
index e60c3c0b9c..7125f5a8e3 100644
--- a/time/tzfile.c
+++ b/time/tzfile.c
@@ -302,14 +302,14 @@  __tzfile_read (const char *file)
      The piece-wise allocations from buf below verify that no
      overflow/wraparound occurred in these computations.
 
-     The order of the suballocations is important for alignment
-     purposes.  __time64_t outside a struct may require more alignment
-     then inside a struct on some architectures, so it must come
-     first. */
-  _Static_assert (__alignof (__time64_t) >= __alignof (struct leap),
-		  "alignment of __time64_t");
-  _Static_assert (__alignof (struct leap) >= __alignof (struct ttinfo),
-		  "alignment of struct leap");
+     The order of the suballocations is important for alignment purposes.
+     The calculation of ALIGNMENT_SLOP assumes arrays are allocated in
+     element order __time_64_t, struct leap, struct ttinfo.  */
+  size_t alignment_slop =
+    ((__alignof__ (struct leap) <= __alignof__ (__time64_t)
+      ? 0 : __alignof__ (struct leap) - __alignof__ (__time64_t))
+     + (__alignof__ (struct ttinfo) <= __alignof__ (struct leap)
+	? 0 : __alignof__ (struct ttinfo) - __alignof__ (struct leap)));
   struct alloc_buffer buf;
   {
     size_t total_size, product;
@@ -322,6 +322,7 @@  __tzfile_read (const char *file)
     v |= ckd_add (&total_size, total_size, num_transitions); /* type_idxs */
     v |= ckd_add (&total_size, total_size, chars); /* zone_names */
     v |= ckd_add (&total_size, total_size, tzspec_size);
+    v |= ckd_add (&total_size, total_size, alignment_slop);
     if (v)
       goto lose;
     transitions = malloc (total_size);