JDK-8304291 : [AIX] Broken build after JDK-8301998
  • Type: Bug
  • Component: client-libs
  • Sub-Component: 2d
  • Affected Version: 21
  • Priority: P3
  • Status: Resolved
  • Resolution: Fixed
  • OS: aix
  • Submitted: 2023-03-15
  • Updated: 2023-06-14
  • Resolved: 2023-04-20
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
JDK 11 JDK 17 JDK 20 JDK 21
11.0.20Fixed 17.0.8Fixed 20.0.2Fixed 21 b20Fixed
Related Reports
Duplicate :  
JDK-8304321 - AIX build fails after JDK-8301998 Update HarfBuzz to 7.0.1
Relates :  
JDK-8301998 - Update HarfBuzz to 7.0.1
Description
Harfbuzz version 7.0.1 does not appear to compile on AIX. Bumping to 7.1.0 fixes one issue, but another issue is present after fixing that.

The first build error is:
Error while processing /home/hotspot/openjdk/jdk-deepa/src/java.desktop/share/native/libharfbuzz/hb-aat-map.cc.
In file included from /home/hotspot/openjdk/jdk-deepa/src/java.desktop/share/native/libharfbuzz/hb-buffer.cc:30:
In file included from /home/hotspot/openjdk/jdk-deepa/src/java.desktop/share/native/libharfbuzz/hb-buffer.hh:33:
In file included from /home/hotspot/openjdk/jdk-deepa/src/java.desktop/share/native/libharfbuzz/hb.hh:513:
/home/hotspot/openjdk/jdk-deepa/src/java.desktop/share/native/libharfbuzz/hb-algs.hh:882:10: error: 1540-2993 The builtin "__builtin_mul_overflow" is not supported.
  return __builtin_mul_overflow (count, size, result);

Then:

=== Output from failing command(s) repeated here ===
* For target support_native_java.desktop_libfontmanager_hb-subset.o:
In file included from /home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-subset.cc:27:
In file included from /home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/OT/Color/COLR/../../../hb.hh:522:
/home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-iter.hh:372:23: error: no matching function for call to object of type 'struct (anonymous struct at /home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-algs.hh:414:1)'
            decltype (hb_get (hb_declval (Proj), *hb_declval (Iter)))>
                      ^~~~~~
/home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-map.hh:346:5: note: in instantiation of template class 'hb_map_iter_t<hb_filter_iter_t<hb_array_t<hb_hashmap_t<hb_array_t<const char>, unsigned int, true>::item_t>, bool (hb_hashmap_t<hb_array_t<const char>, unsigned int, true>::item_t::*)() const, (anonymous struct at /home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-algs.hh:202:1) &, nullptr>, const char *, hb_function_sortedness_t::NOT_SORTED, nullptr>' requested here
    + keys_ref ()
    ^
/home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-ot-post-table-v2subset.hh:81:44: note: in instantiation of template class 'hb_hashmap_t<hb_array_t<const char>, unsigned int, true>' requested here
  hb_hashmap_t<hb_bytes_t, uint32_t, true> glyph_name_to_new_index;
                                           ^
/home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-algs.hh:440:3: note: candidate template ignored: substitution failure [with Proj = const char *, Val = hb_hashmap_t<hb_array_t<const char>, unsigned int, true>::item_t &]: no matching member function for call to 'impl'
  operator () (Proj&& f, Val &&v) const HB_AUTO_RETURN
  ^
In file included from /home/hotspot/openjdk/jdk-tyler/src/java.desktop/share/native/libharfbuzz/hb-subset.cc:27:
   ... (rest of output omitted)

* All command lines available in /home/hotspot/openjdk/jdk-tyler/build/aix-ppc64-server-fastdebug/make-support/failure-logs.
=== End of repeated output ===
Comments
A pull request was submitted for review. URL: https://git.openjdk.org/jdk11u/pull/74 Date: 2023-06-09 12:41:51 +0000
09-06-2023
Another PR here: https://github.com/openjdk/jdk11u/pull/74
09-06-2023
Thanks Goetz! Will do!
09-06-2023
A pull request was submitted for review. URL: https://git.openjdk.org/jdk11u-dev/pull/1940 Date: 2023-06-09 12:18:50 +0000
09-06-2023
[~adfarley] This is broken in 11.0.20, too, right? Don't you want to open a PR against jdk11u and file a jdk11u-critical-request? Anyways, please only label jdk11u-fix-request if the backport is ready to be pushed!
06-06-2023
Fix Request [11u]: The __builtin_mul_overflow error has been seen while building JDK11u on AIX. Since the line numbers relevant to the change appear the same as on later versions, I'd like to request a backport of this fix, please.
05-06-2023
In head, we should probably replace !AIX by a check for xlc > 16?
04-05-2023
A pull request was submitted for review. URL: https://git.openjdk.org/jdk17u-dev/pull/1320 Date: 2023-05-03 19:32:27 +0000
03-05-2023
fix request [20u, 17u]: This change is required in order to build after the Harfbuzz upgrade (JDK-8304321). The risk is medium, because this patch may need to be re-applied each time the Harfbuzz library upgrade is performed (it is unlikely that such a change will be made to the Harfbuzz library upstream). Testing: T1 is as expected, test/hotspot/jtreg/serviceability/jvmti are all passing. Patch applied cleanly.
03-05-2023
A pull request was submitted for review. URL: https://git.openjdk.org/jdk20u/pull/57 Date: 2023-04-20 20:47:40 +0000
20-04-2023
Changeset: 310aa934 Author: Matthias Baesken <mbaesken@openjdk.org> Date: 2023-04-20 07:07:00 +0000 URL: https://git.openjdk.org/jdk/commit/310aa9347861922af5f0311e9e93a5f49dee6adc
20-04-2023
A pull request was submitted for review. URL: https://git.openjdk.org/jdk/pull/13529 Date: 2023-04-19 10:38:19 +0000
19-04-2023
> That '__builtin_mul_overflow' is fixed in Harfbuzz release 7.1. > Would you upgrade, or just add a patch to libharfbuzz with the change? I would not upgrade now to 7.1 but just add the patch. In a couple of months we will probably go to xlc17 and might remove the patch afterwards. Btw that is the patch, latest jdk nightbuild worked , could you try it as well please? diff --git a/src/java.desktop/share/native/libharfbuzz/hb-algs.hh b/src/java.desktop/share/native/libharfbuzz/hb-algs.hh index 28dc036a515..3c96884ff08 100644 --- a/src/java.desktop/share/native/libharfbuzz/hb-algs.hh +++ b/src/java.desktop/share/native/libharfbuzz/hb-algs.hh @@ -876,10 +876,13 @@ static inline bool hb_unsigned_mul_overflows (unsigned int count, unsigned int size, unsigned *result = nullptr) { #if (defined(__GNUC__) && (__GNUC__ >= 4)) || (defined(__clang__) && (__clang_major__ >= 8)) +/* avoid on old xlc16 clang */ +#if !defined(AIX) unsigned stack_result; if (!result) result = &stack_result; return __builtin_mul_overflow (count, size, result); +#endif #endif if (result) diff --git a/src/java.desktop/share/native/libharfbuzz/hb-subset.cc b/src/java.desktop/share/native/libharfbuzz/hb-subset.cc index a8161a32551..f6f37527733 100644 --- a/src/java.desktop/share/native/libharfbuzz/hb-subset.cc +++ b/src/java.desktop/share/native/libharfbuzz/hb-subset.cc @@ -24,6 +24,7 @@ * Google Author(s): Garret Rieger, Rod Sheeter, Behdad Esfahbod */ +#if !defined(AIX) #include "hb.hh" #include "hb-open-type.hh" @@ -631,3 +632,6 @@ end: /* If NO_VISIBILITY, libharfbuzz has this. */ #include "hb-ot-name-language-static.hh" #endif + +// AIX +#endif
06-04-2023
That's terrific news :-) I agree with your conclusion [about hb-subset.cc], but I hadn't thought of excluding it from the build. That '__builtin_mul_overflow' is fixed in Harfbuzz release 7.1. Would you upgrade, or just add a patch to libharfbuzz with the change?
05-04-2023
Hi Tyler, it seems to me that the "bad" file src/java.desktop/share/native/libharfbuzz/hb-subset.cc that triggers the compile errors with xlC16 can be excluded from compilation in the OpenJDK. At least I see no compile/link errors when doing so on AIX; maybe we do not use the hb-font subsetting at all in OpenJDK ? Do I miss something? Should we do so, would also reduce compile time a little bit. With hb-subset.cc excluded and a small adjustment for the "__builtin_mul_overflow" in hb-algs.hh, the build succeeds with xlC16.
05-04-2023
The first "__builtin_mul_overflow" is not supported - issue can be easily handled by adjusting the related ifdef at the point where compilation breaks. The second issue "error: no matching function for call ..." is more tricky, but shows up, according to harfbuzz issue 4138, as well on macOS when ancient clang versions are used. So there is a very good chance that with a new clang it works nicely; clang 13 on macOS works for us. xlc 16.1.X is still based on the very old clang 4.0.X . xlc 17.1.1 seems to be based on clang 15.0.0, so this might be an option to get around this problem on AIX .
22-03-2023
Hi Martin, Good sleuthing. Your linked issue certainly means that xlclang must be updated to include fixpack 2.
17-03-2023
Is one of the issues related to https://www.ibm.com/support/pages/apar/IJ14858 ?
16-03-2023
Second issue open at Harfbuzz repo: https://github.com/harfbuzz/harfbuzz/issues/4138
15-03-2023