Re: java: switch to icedtea-8 as default JDK

2017-12-04 16:44 GMT+01:00 Gábor Boskovits <address@hidden>:

Now that this problem around glibc is resolved, I think I will do some history rewrite, so that these reverts, reverting the revert.... does not show up.
I 'm also willing to rename the branch to have wip in the name, as this seems to be standard for longer runnig parts. WDYT?

2017-12-04 15:19 GMT+01:00 Gábor Boskovits <address@hidden>:
Ok, so it seems, that the latest glibc update was made to fix the icu4c thing.
It just broke the locales.
Fix for that has just been pushed, making core-updates usable again.
I pushed to my icedtea branch also.
I'm now building it to see what we have now.

2017-12-04 13:59 GMT+01:00 Gábor Boskovits <address@hidden>:
Upstream corrected the issue in commit 6913ad65e00bb32417ad39c41d292b976171e27e.

It is not yet included in 2.26.

The patch is basically the same as I proposed.

I think we should get the upstream version.

2017-12-04 13:33 GMT+01:00 Gábor Boskovits <address@hidden>:
I think it should look like this:

--- cdefs.h 2017-12-04 13:27:55.640000000 +0100
+++ cdefs.new.h 2017-12-04 13:31:53.820000000 +0100
@@ -1,4 +1,4 @@
-/* Copyright (C) 1992-2017 Free Software Foundation, Inc.
+y/* Copyright (C) 1992-2017 Free Software Foundation, Inc.
This file is part of the GNU C Library.

The GNU C Library is free software; you can redistribute it and/or
@@ -471,10 +471,10 @@
testing __STDC_VERSION__ for generic selection support.
On the other hand, Clang also defines __GNUC__, so a clang-specific
check is required to enable the use of generic selection. */
-#if __GNUC_PREREQ (4, 9) \
+#if !defined __cplusplus && (__GNUC_PREREQ (4, 9) \
|| __glibc_clang_has_extension (c_generic_selections) \
|| (!defined __GNUC__ && defined __STDC_VERSION__ \
- && __STDC_VERSION__ >= 201112L)
+ && __STDC_VERSION__ >= 201112L))
# define __HAVE_GENERIC_SELECTION 1
#else
# define __HAVE_GENERIC_SELECTION 0

2017-12-04 13:21 GMT+01:00 Gábor Boskovits <address@hidden>:
We also have this in bits/floatn.h, but this seems right.

#if (defined __x86_64__ \
? __GNUC_PREREQ (4, 3) \
: (defined __GNU__ ? __GNUC_PREREQ (4, 5) : __GNUC_PREREQ (4, 4)))
# define __HAVE_FLOAT128 1
#else
# define __HAVE_FLOAT128 0
#endif

/* Defined to 1 if __HAVE_FLOAT128 is 1 and the type is ABI-distinct
from the default float, double and long double types in this glibc. */
#if __HAVE_FLOAT128
# define __HAVE_DISTINCT_FLOAT128 1
#else
# define __HAVE_DISTINCT_FLOAT128 0
#endif

I think the problem is caused by HAVE_GENERIC_SELECTION defined when we are in C++.

2017-12-04 13:18 GMT+01:00 Gábor Boskovits <address@hidden>:
We have this in cdefs.h:
#if __GNUC_PREREQ (4, 9) \
|| __glibc_clang_has_extension (c_generic_selections) \
|| (!defined __GNUC__ && defined __STDC_VERSION__ \
&& __STDC_VERSION__ >= 201112L)
# define __HAVE_GENERIC_SELECTION 1
#else
# define __HAVE_GENERIC_SELECTION 0
#endif

This does not seem right.

We should not have this defined when compiling c++.
WDYT?

2017-12-04 13:03 GMT+01:00 Gábor Boskovits <address@hidden>:
This minimal file reproduces the problem in icu4c 60.1 in guix environment icu4c on current core-updates.

#include <cmath>

int main(int argc, char **argv)
{
std::signbit(1);
return 0;
}

It seems, that it is some problem with cmath.h

/gnu/store/2wzijwwgpcdb1msmn3wgq9zmrbrkpx2h-gcc-5.5.0/include/c++/cmath:44:0

The definition is in glibc 2.26 math.h, looks like this:

#elif __HAVE_DISTINCT_FLOAT128
# if __HAVE_GENERIC_SELECTION
# define __MATH_TG(TG_ARG, FUNC, ARGS) \
_Generic ((TG_ARG), \
float: FUNC ## f ARGS, \
default: FUNC ARGS, \
long double: FUNC ## l ARGS, \
_Float128: FUNC ## f128 ARGS)

2017-12-03 23:25 GMT+01:00 Gábor Boskovits <address@hidden>:
It seems, that the fix is already included in the 60.1 release for the xlocal problem. I think we can ignore that.

2017-12-03 23:20 GMT+01:00 Gábor Boskovits <address@hidden>:
It seems, that they issued that already:
http://bugs.icu-project.org/trac/changeset/40603

2017-12-03 23:08 GMT+01:00 Gábor Boskovits <address@hidden>:
Ok, I found this one:
https://sourceware.org/glibc/wiki/Release/2.26#Removal_of_.27xlocale.h.27

It seems, that xlocale.h should not be used as an include, and glibc removed that.

So this is ok.

The question is if this issue was addressed by icu4c, so we can just go to the next step, or should we do something about this.

2017-12-03 23:04 GMT+01:00 Gábor Boskovits <address@hidden>:
I've found that simply reverting the update to icu4c 60 commit brings up the issue you just mentioned, with missing xlocale.h.
So it seems, that the one you found introduced the xlocale problem, while 4e080fbb0bb73e4444181599676f4e1ef5fdc2ba introduces the new error.
I think we should deal with the one you find first, the see what's with this newer one.

We also have a long running discussion on a bug introduced by a later glibc update, ee3ebf1a357bd4eb36a2fa1790a7b549cffb305a which broke a lot of packages. Might these be somehow related?
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=29537

2017-12-03 22:46 GMT+01:00 Chris Marusich <address@hidden>:
Gábor Boskovits <address@hidden> writes:

> Any news on this icu4c thing?

After about 2 days of running git bisect, my computer has informed me
that the first bad commit is 67d527e35e367c9e9e89ec01cda2ce32cabd2d89.
This is the first commit where icu4c fails to build on core-updates.
The build failure at this commit is as follows:

--8<---------------cut here---------------start------------->8---
g++ ... decimfmt.cpp
g++ ... decimalformatpattern.cpp
g++ ... dcfmtsym.cpp
g++ ... digitlst.cpp
g++ ... fmtable_cnv.cpp
digitlst.cpp:67:24: fatal error: xlocale.h: No such file or directory
compilation terminated.
*** Failed compilation command follows: ----------------------------------------------------------
g++ -D_REENTRANT -DU_HAVE_ELF_H=1 -DU_HAVE_ATOMIC=1 -DU_HAVE_STRTOD_L=1 -I. -I../common -DU_ATTRIBUTE_DEPRECATED= -DU_I18N_IMPLEMENTATION -O2 -W -Wall -pedantic -Wpointer-arith -Wwrite-strings -Wno-long-long --std=c++0x -c -DPIC -fPIC -o digitlst.o digitlst.cpp
--- ( rebuild with "make VERBOSE=1 all" to show all parameters ) --------
make[1]: *** [../config/mh-linux:51: digitlst.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/tmp/guix-build-icu4c-58.2.drv-0/icu/source/i18n'
make: *** [Makefile:143: all-recursive] Error 2
phase `build' failed after 59.0 seconds
builder for `/gnu/store/mdd6glhc0dg65y4wd11y0b7sbky9cwv6-icu4c-58.2.drv' failed with exit code 1
@ build-failed /gnu/store/mdd6glhc0dg65y4wd11y0b7sbky9cwv6-icu4c-58.2.drv - 1 builder for `/gnu/store/mdd6glhc0dg65y4wd11y0b7sbky9cwv6-icu4c-58.2.drv' failed with exit code 1
guix build: error: build failed: build of `/gnu/store/mdd6glhc0dg65y4wd11y0b7sbky9cwv6-icu4c-58.2.drv' failed
--8<---------------cut here---------------end--------------->8---

Note that this is NOT the same error as the one that occurs at commit
d6adba786cf2ab1b26ff083928b64262281ff106 (which is the commit on
core-updates from which Gábor's branch change-default-icedtea-8 branch
begins). That error is:

--8<---------------cut here---------------start------------->8---
g++ ... number_decimalquantity.cpp

g++ ... number_decimfmtprops.cpp
number_decimalquantity.cpp: In member function ‘icu_60::number::impl::DecimalQuantity& icu_60::number::impl::DecimalQuantity::setToDouble(double)’:
number_decimalquantity.cpp:333:9: error: ‘_Generic’ is not a member of ‘std’
if (std::signbit(n) != 0) {
^
In file included from /gnu/store/nz2m4gdvgzcrkqa4xwv360iskh7syj7i-gcc-5.5.0/include/c++/cmath:44:0,
from number_decimalquantity.cpp:9:
number_decimalquantity.cpp:333:14: error: expected primary-_expression_ before ‘float’
if (std::signbit(n) != 0) {
^
number_decimalquantity.cpp:333:14: error: expected primary-_expression_ before ‘default’
if (std::signbit(n) != 0) {
^
number_decimalquantity.cpp:333:14: error: expected primary-_expression_ before ‘long’
if (std::signbit(n) != 0) {
^
number_decimalquantity.cpp:333:14: error: expected primary-_expression_ before ‘:’ token
if (std::signbit(n) != 0) {
^
number_decimalquantity.cpp:337:9: error: ‘__builtin_isnan’ is not a member of ‘std’
if (std::isnan(n) != 0) {
^
number_decimalquantity.cpp:337:9: note: suggested alternative:
<built-in>: note: ‘__builtin_isnan’
number_decimalquantity.cpp:339:16: error: ‘__builtin_isfinite’ is not a member of ‘std’
} else if (std::isfinite(n) == 0) {
^
number_decimalquantity.cpp:339:16: note: suggested alternative:
<built-in>: note: ‘__builtin_isfinite’

*** Failed compilation command follows: ----------------------------------------------------------
g++ -D_REENTRANT -DU_HAVE_ELF_H=1 -DU_HAVE_ATOMIC=1 -DU_HAVE_STRTOD_L=1 -DU_HAVE_XLOCALE_H=0 -I. -I../common -DU_ATTRIBUTE_DEPRECATED= -DU_I18N_IMPLEMENTATION -O2 -W -Wall -pedantic -Wpointer-arith -Wwrite-strings -Wno-long-long -std=c++11 -c -DPIC -fPIC -o number_decimalquantity.o number_decimalquantity.cpp
--- ( rebuild with "make VERBOSE=1 all" to show all parameters ) --------
make[1]: *** [../config/mh-linux:51: number_decimalquantity.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/tmp/guix-build-icu4c-60.1.drv-0/icu/source/i18n'
make: *** [Makefile:149: all-recursive] Error 2
phase `build' failed after 122.2 seconds
builder for `/gnu/store/8s6q5cll4knh7y0wfrbjqs2dai0x4sm2-icu4c-60.1.drv' failed with exit code 1
@ build-failed /gnu/store/8s6q5cll4knh7y0wfrbjqs2dai0x4sm2-icu4c-60.1.drv - 1 builder for `/gnu/store/8s6q5cll4knh7y0wfrbjqs2dai0x4sm2-icu4c-60.1.drv' failed with exit code 1
guix build: error: build failed: build of `/gnu/store/8s6q5cll4knh7y0wfrbjqs2dai0x4sm2-icu4c-60.1.drv' failed
--8<---------------cut here---------------end--------------->8---

The fact that these errors are different suggests that more than one
problem may have been introduced...

I haven't tried any more recent commits on core-updates. I am not sure
what the best way is to debug this, so I would appreciate any advice
anyone has.

I hope that information helps. Hopefully, together, we can untangle
this!

--
Chris

From:	Gábor Boskovits
Subject:	Re: java: switch to icedtea-8 as default JDK
Date:	Mon, 4 Dec 2017 17:34:53 +0100