RFR: 8266545: 8261169 broke Harfbuzz build with gcc 7 and 8

Aleksey Shipilev shade at openjdk.java.net
Thu May 6 06:19:49 UTC 2021


On Wed, 5 May 2021 07:54:20 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

> Harfbuzz upgrade broke Linux x64 build on older gccs. For details see JBS issue.
> 
> I fixed the issue in the harfbuzz sources, but I am not sure of the policy here. Do we modify the harfbuzz sources or leave them untouched? Advice is welcome.
> 
> The patch fixes linux x64 fastdebug and opt build for me.

FWIW, current jdk master builds fine with gcc 9.3.0. It fails with gcc 6.3.0 to me.

I am unaware of the policy of changing the Harfbuzz sources too. If we are changing the Harfbuzz sources, I prefer the multi-line form you did in Harfbuzz PR, though: https://github.com/harfbuzz/harfbuzz/pull/2973 -- it would also simplify the merge. 

But there is an alternative, disable the warning and then wait for Harfbuzz fix to drop:


diff --git a/make/modules/java.desktop/lib/Awt2dLibraries.gmk b/make/modules/java.desktop/lib/Awt2dLibraries.gmk
index ff5fa00c720..627aa51ebdf 100644
--- a/make/modules/java.desktop/lib/Awt2dLibraries.gmk
+++ b/make/modules/java.desktop/lib/Awt2dLibraries.gmk
@@ -465,7 +465,7 @@ else
 
    HARFBUZZ_DISABLED_WARNINGS_gcc := type-limits missing-field-initializers strict-aliasing
    HARFBUZZ_DISABLED_WARNINGS_CXX_gcc := reorder delete-non-virtual-dtor strict-overflow \
-        maybe-uninitialized class-memaccess unused-result
+        maybe-uninitialized class-memaccess unused-result extra
    HARFBUZZ_DISABLED_WARNINGS_clang := unused-value incompatible-pointer-types \
         tautological-constant-out-of-range-compare int-to-pointer-cast \
         undef missing-field-initializers range-loop-analysis \

-------------

PR: https://git.openjdk.java.net/jdk/pull/3873



More information about the build-dev mailing list