1862601 - build failure using --system-icu and ICU 74.1

Status: NEW

(Core :: Internationalization, defect, P3)

Description

[Landry Breuil (:gaston)]

building with --with-system-icu on OpenBSD, with a work-in-progress update to ICU 74.1, the build fails:

In file included from Unified_cpp_intl_lwbrk0.cpp:2: 
/usr/obj/ports/firefox-120.0beta5/firefox-120.0/intl/lwbrk/LineBreaker.cpp:453:3: error: static_assert failed due to requirement 'U_LB_COUNT == mozilla::ArrayLength(sUnicodeLineBreakToClass)' "Gecko vs ICU LineBreak class mismatch"
static_assert(U_LB_COUNT == mozilla::ArrayLength(sUnicodeLineBreakToClass),
^             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.  

Comment 1

[Landry Breuil (:gaston)]

icu4c 74.1 has a different U_LB_COUNT than 73.2

/usr/local/include/unicode/uchar.h:    U_LB_COUNT = 43
/usr/obj/ports/icu4c-74_1-src/icu/source/common/unicode/uchar.h:    U_LB_COUNT = 48

maybe the array in https://searchfox.org/mozilla-central/source/intl/lwbrk/LineBreaker.cpp#408 should be updated ?

Comment 2

[Landry Breuil (:gaston)]

afaict this comes from https://github.com/unicode-org/icu/commit/2e45e6ec0e84a1c01812015a254ea31b286316fb#diff-a281b7b66425b3bd279e98797c80c2067a7127dc22792686dfdada157ff5258dL2441 (via https://github.com/unicode-org/unicodetools/pull/422 for the source ?)

Comment 3

[André Bargull [:anba]]

Yes, sUnicodeLineBreakToClass needs to be updated for ICU 74 (bug 1859752). The try-push for the ICU 74 release candidate has this patch:

--- a/intl/lwbrk/LineBreaker.cpp
+++ b/intl/lwbrk/LineBreaker.cpp
@@ -443,17 +443,23 @@ static int8_t GetClass(uint32_t u, LineB
       /* JT = 34,                           [JT] */ CLASS_CHARACTER,
       /* JV = 35,                           [JV] */ CLASS_CHARACTER,
       /* CLOSE_PARENTHESIS = 36,            [CP] */ CLASS_CLOSE_LIKE_CHARACTER,
       /* CONDITIONAL_JAPANESE_STARTER = 37, [CJ] */ CLASS_CLOSE,
       /* HEBREW_LETTER = 38,                [HL] */ CLASS_CHARACTER,
       /* REGIONAL_INDICATOR = 39,           [RI] */ CLASS_CHARACTER,
       /* E_BASE = 40,                       [EB] */ CLASS_BREAKABLE,
       /* E_MODIFIER = 41,                   [EM] */ CLASS_CHARACTER,
-      /* ZWJ = 42,                          [ZWJ]*/ CLASS_CHARACTER};
+      /* ZWJ = 42,                          [ZWJ]*/ CLASS_CHARACTER,
+      /* AKSARA = 43,                       [AK] */ CLASS_CHARACTER,
+      /* AKSARA_PREBASE = 44,               [AP] */ CLASS_CHARACTER,
+      /* AKSARA_START = 45,                 [AS] */ CLASS_CHARACTER,
+      /* VIRAMA_FINAL = 46,                 [VF] */ CLASS_CHARACTER,
+      /* VIRAMA = 47,                       [VI] */ CLASS_CHARACTER,
+  };
 
   static_assert(U_LB_COUNT == mozilla::ArrayLength(sUnicodeLineBreakToClass),
                 "Gecko vs ICU LineBreak class mismatch");
 
   auto cls = GetLineBreakClass(u);
   MOZ_ASSERT(cls < mozilla::ArrayLength(sUnicodeLineBreakToClass));
 
   // Overrides based on rules for the different line-break values given in

Comment 4

[Landry Breuil (:gaston)]

(In reply to André Bargull [:anba] from comment #3)

Yes, sUnicodeLineBreakToClass needs to be updated for ICU 74 (bug 1859752). The try-push for the ICU 74 release candidate has this patch:

--- a/intl/lwbrk/LineBreaker.cpp
+++ b/intl/lwbrk/LineBreaker.cpp
@@ -443,17 +443,23 @@ static int8_t GetClass(uint32_t u, LineB
       /* JT = 34,                           [JT] */ CLASS_CHARACTER,
       /* JV = 35,                           [JV] */ CLASS_CHARACTER,
       /* CLOSE_PARENTHESIS = 36,            [CP] */ CLASS_CLOSE_LIKE_CHARACTER,
       /* CONDITIONAL_JAPANESE_STARTER = 37, [CJ] */ CLASS_CLOSE,
       /* HEBREW_LETTER = 38,                [HL] */ CLASS_CHARACTER,
       /* REGIONAL_INDICATOR = 39,           [RI] */ CLASS_CHARACTER,
       /* E_BASE = 40,                       [EB] */ CLASS_BREAKABLE,
       /* E_MODIFIER = 41,                   [EM] */ CLASS_CHARACTER,
-      /* ZWJ = 42,                          [ZWJ]*/ CLASS_CHARACTER};
+      /* ZWJ = 42,                          [ZWJ]*/ CLASS_CHARACTER,
+      /* AKSARA = 43,                       [AK] */ CLASS_CHARACTER,
+      /* AKSARA_PREBASE = 44,               [AP] */ CLASS_CHARACTER,
+      /* AKSARA_START = 45,                 [AS] */ CLASS_CHARACTER,
+      /* VIRAMA_FINAL = 46,                 [VF] */ CLASS_CHARACTER,
+      /* VIRAMA = 47,                       [VI] */ CLASS_CHARACTER,
+  };
 
   static_assert(U_LB_COUNT == mozilla::ArrayLength(sUnicodeLineBreakToClass),
                 "Gecko vs ICU LineBreak class mismatch");
 
   auto cls = GetLineBreakClass(u);
   MOZ_ASSERT(cls < mozilla::ArrayLength(sUnicodeLineBreakToClass));
 
   // Overrides based on rules for the different line-break values given in

thanks, i had more or less the same wip patch locally, will confirm if it fixes the build and if runtime is still ok.

Comment 5

[Landry Breuil (:gaston)]

runtime seems ok using 120.0b5 with https://hg.mozilla.org/try/rev/d5f3b0c4f08a426ce00a153c04e177eecb6820e2 applied and built against systemwide icu 74.1.

Comment 6

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:m_kato, could you have a look please?

For more information, please visit BugBot documentation.

Comment 7

[tt_1]

the 115esr branch is also affected

Comment 8

[Joonas Niilola]

Attachment: firefox-120.0:20231122-063209.log.xz

With the patch build works with unified build, but with --disable-unified-build the build fails to:

intl/components/src/TimeZone.cpp:345:3: error: use of undeclared identifier 'MOZ_TRY'

Full build log attached as compressed.

Comment 9

[Joonas Niilola]

With the patch --disable-unified-build is broken, but a unified build works fine. I get: