[ANN] GCC 14.2.0-3 (aarch64, macOS)

[ANN] GCC 14.2.0-3 (aarch64, macOS)

[Simon Wright]

Available at [1]. Included tools at v25.0.0.

[1] https://github.com/simonjwright/distributing-gcc/releases/tag/gcc-14.2.0-3-aarch64

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Bill Findlay]

On 17 Nov 2024, Simon Wright wrote
(in article <lyserpq5wi.fsf@pushface.org>):

> Available at [1]. Included tools at v25.0.0.
>
> [1] https://github.com/simonjwright/distributing-gcc/releases/tag/gcc-14.2.0-3
> -aarch64

Hi Simon, many thanks for that.

I note that 14.2.0-3 has the same issue as all previous 14.2 versions:
a spurious error message on a simple string concatenation,
which appears only when link-time optimization is called for:

gnatmake -aI../Source -aO../Build -funwind-tables -gnatl12j96 -gnatw.e 
-gnatwD -gnatwH -gnatwP -gnatwT -gnatw.W -gnatw.B -gnatwC -gnatw.u -gnatyO 
-gnatw.Y -gnatw.N -fdata-sections -ffunction-sections -O3 -flto askance 
-bargs -static -Sin -largs -Wl,-dead_strip -Wl,-dead_strip -largs -flto
...
gnatbind -aI../Source -aO../Build -static -Sin -x askance.ali
gnatlink askance.ali -funwind-tables -fdata-sections -ffunction-sections -O3 
-flto -Wl,-dead_strip -Wl,-dead_strip -flto
lto-wrapper: warning: using serial compilation of 4 LTRANS jobs
lto-wrapper: note: see the '-flto' option documentation for more information
/Users/wf/KDF9/emulation/Source/posix.adb: In function 'posix__output_line':
/Users/wf/KDF9/emulation/Source/posix.adb:277:49: warning: '__builtin_memcpy' 
writing between 1 and 2147483647 bytes into a region of size 0 
[-Wstringop-overflow=]
277 | message_line : constant String := message &  NL;
| ^

This is in the routine:

procedure output_line (message : in String) is
message_line : String := message &  NL;
begin
output(message_line);
end output_line;

where:

NL : constant String := OS_specifics.EOL;

in the package spec., and:

package body OS_specifics is

function EOL
return String
is (1 =>  Character'Val(16#0A#));
...

Changing the complained-of line to:

message_line : String := message &  (NL &  "");

makes the warning go away.

-- 
Bill Findlay

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[moi]

On 21/11/2024 15:30, Bill Findlay wrote:
> On 17 Nov 2024, Simon Wright wrote
> (in article <lyserpq5wi.fsf@pushface.org>):
> 
>> Available at [1]. Included tools at v25.0.0.
>>
>> [1] https://github.com/simonjwright/distributing-gcc/releases/tag/gcc-14.2.0-3
>> -aarch64
> 
> Hi Simon, many thanks for that.
> 
> I note that 14.2.0-3 has the same issue as all previous 14.2 versions:
> a spurious error message on a simple string concatenation,
> which appears only when link-time optimization is called for:
...
That was garbled by the newsreader! Should be:

> lto-wrapper: note: see the '-flto' option documentation for more information
> /Users/wf/KDF9/emulation/Source/posix.adb: In function 'posix__output_line':
> /Users/wf/KDF9/emulation/Source/posix.adb:277:49: warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes into a region of size 0
> [-Wstringop-overflow=]
> 277 | message_line : constant String := message &  NL;
>     |                                           ^

with the ^ pointing to the &.

I forgot to say that I'm running up-to-date Sequoia.

-- 
Bill F.

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Simon Wright]

I can't reproduce.

The only change made to the 14.2.0 compiler release are to do with
fixincludes vs SDK16.

-1:     gcc-14.2-darwin-r1
-2, -3: gcc-14.2-darwin-r2

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[moi]

On 21/11/2024 20:40, Simon Wright wrote:
> I can't reproduce.
> 
> The only change made to the 14.2.0 compiler release are to do with
> fixincludes vs SDK16.
> 
> -1:     gcc-14.2-darwin-r1
> -2, -3: gcc-14.2-darwin-r2

I'm not surpised, it seems to be very dependent on the exact context.
I could not reproduce it in a simple example.
It has been present in all 14* compilers, but it is absent from 13.2.0.
-- 
Bill F.

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Björn Persson]

Bill Findlay wrote:
> warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes
> into a region of size 0 [-Wstringop-overflow=]

I've seen many occurrences of that bogus warning in Fedora, so it's not
unique to Simon's build. Thus the place to report it would
be https://gcc.gnu.org/bugzilla/.

Björn Persson

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[moi]

On 22/11/2024 10:02, Björn Persson wrote:
> Bill Findlay wrote:
>> warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes
>> into a region of size 0 [-Wstringop-overflow=]
> 
> I've seen many occurrences of that bogus warning in Fedora, so it's not
> unique to Simon's build. Thus the place to report it would
> be https://gcc.gnu.org/bugzilla/.
> 
> Björn Persson
> 
Do you get it only in 14.2.0-*, Björn?
-- 
Bill F.

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Simon Wright]

moi <findlaybill@blueyonder.co.uk> writes:

> On 22/11/2024 10:02, Björn Persson wrote:
>> Bill Findlay wrote:
>>> warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes
>>> into a region of size 0 [-Wstringop-overflow=]
>> I've seen many occurrences of that bogus warning in Fedora, so it's
>> not unique to Simon's build. Thus the place to report it would be
>> https://gcc.gnu.org/bugzilla/.
>> Björn Persson
>> 
> Do you get it only in 14.2.0-*, Björn?

It happens with gcc version 15.0.0 20241102 (experimental) :-(

And with an x86_64-apple-darwin GCC 4.2.0 build (on aarch64, under
Rosetta) I noticed

In function 'disassembly__data_access_name',
    inlined from 'disassembly__the_full_name_of' at /Users/simon/tmp/emulation/Source/disassembly.adb:483:44:
/Users/simon/tmp/emulation/Source/disassembly.adb:452:48: warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes into a region of size 0 overflows the destination [-Wstringop-overflow=]
  452 |       modifier  : constant String  := M_suffix & Q_suffix;
      |                                                ^
/Users/simon/tmp/emulation/Source/disassembly.adb:452:48: note: destination object 'S1471b.315' of size 0
  452 |       modifier  : constant String  := M_suffix & Q_suffix;
      |                                                ^
/Users/simon/tmp/emulation/Source/disassembly.adb:452:48: warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes into a region of size 0 overflows the destination [-Wstringop-overflow=]
  452 |       modifier  : constant String  := M_suffix & Q_suffix;
      |                                                ^
/Users/simon/tmp/emulation/Source/disassembly.adb:452:48: note: destination object 'S1471b.315' of size 0
  452 |       modifier  : constant String  := M_suffix & Q_suffix;
      |                                                ^

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Björn Persson]

moi wrote:
> On 22/11/2024 10:02, Björn Persson wrote:
> > Bill Findlay wrote:  
> >> warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes
> >> into a region of size 0 [-Wstringop-overflow=]  
> > 
> > I've seen many occurrences of that bogus warning in Fedora, so it's not
> > unique to Simon's build. Thus the place to report it would
> > be https://gcc.gnu.org/bugzilla/.
> > 
> > Björn Persson
> >   
> Do you get it only in 14.2.0-*, Björn?

I haven't taken notes of the GCC version. I just concluded that the
warning was obviously wrong, and moved on. I can let you know if I see
it again.

Björn Persson

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Björn Persson]

moi wrote:
> On 22/11/2024 10:02, Björn Persson wrote:
> > Bill Findlay wrote:  
> >> warning: '__builtin_memcpy' writing between 1 and 2147483647 bytes
> >> into a region of size 0 [-Wstringop-overflow=]  
> > 
> > I've seen many occurrences of that bogus warning in Fedora, so it's not
> > unique to Simon's build. Thus the place to report it would
> > be https://gcc.gnu.org/bugzilla/.
> > 
> > Björn Persson
> >   
> Do you get it only in 14.2.0-*, Björn?

It happened today in GCC 14.2.1 (as packaged in Fedora 41), so no, not
only in 14.2.0.

Björn Persson

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Simon Wright]

Björn Persson <Bjorn@rombobjörn.se> writes:

> It happened today in GCC 14.2.1 (as packaged in Fedora 41), so no, not
> only in 14.2.0.

There's no official FSF 14.2.1 release - it may just be like Alire,
which only handles 3 levels, so they call the first packaging of 14.2.0
14.2.1. If you say 'gcc -v' it'll probably say 14.2.0.

Of course I could be completely wrong and Fedora have added lots of
value!

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Keith Thompson]

Simon Wright <simon@pushface.org> writes:
> Björn Persson <Bjorn@rombobjörn.se> writes:
>> It happened today in GCC 14.2.1 (as packaged in Fedora 41), so no, not
>> only in 14.2.0.
>
> There's no official FSF 14.2.1 release - it may just be like Alire,
> which only handles 3 levels, so they call the first packaging of 14.2.0
> 14.2.1. If you say 'gcc -v' it'll probably say 14.2.0.
>
> Of course I could be completely wrong and Fedora have added lots of
> value!

I see some potential for confusion, since there almost certainly will be
an official gcc 14.2.1 release in the near future.  That official
release will include code that's not included in what Fedora calls gcc
14.2.1.  It's a point release, so I wouldn't expect substantial changes,
but still, I think Fedora should use a different naming scheme.

-- 
Keith Thompson (The_Other_Keith) Keith.S.Thompson+u@gmail.com
void Void(void) { Void(); } /* The recursive call of the void */

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Simon Wright]

Keith Thompson <Keith.S.Thompson+u@gmail.com> writes:

> Simon Wright <simon@pushface.org> writes:
>> Björn Persson <Bjorn@rombobjörn.se> writes:
>>> It happened today in GCC 14.2.1 (as packaged in Fedora 41), so no, not
>>> only in 14.2.0.
>>
>> There's no official FSF 14.2.1 release - it may just be like Alire,
>> which only handles 3 levels, so they call the first packaging of 14.2.0
>> 14.2.1. If you say 'gcc -v' it'll probably say 14.2.0.
>>
>> Of course I could be completely wrong and Fedora have added lots of
>> value!
>
> I see some potential for confusion, since there almost certainly will be
> an official gcc 14.2.1 release in the near future.  That official
> release will include code that's not included in what Fedora calls gcc
> 14.2.1.  It's a point release, so I wouldn't expect substantial changes,
> but still, I think Fedora should use a different naming scheme.

I have this vague memory that 14.2.1 is a draft for 14.3.0, and that GCC
never has .1 _releases_. In the same way that 15.0.0 migrates to 15.0.1
before the 15.1.0 release.

Yes, see [1], section "Version Numbering Scheme for GCC 5 and Up".

Also, in gcc-mirror on github, gcc/BASE-VER in the releases/gcc-14
branch [2] holds 14.2.1.

[1] https://gcc.gnu.org/develop.html
[2] https://github.com/gcc-mirror/gcc/blob/releases/gcc-14/gcc/BASE-VER

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Björn Persson]

Keith Thompson wrote:
> Simon Wright <simon@pushface.org> writes:
> > Björn Persson <Bjorn@rombobjörn.se> writes:  
> >> It happened today in GCC 14.2.1 (as packaged in Fedora 41), so no, not
> >> only in 14.2.0.  
> >
> > There's no official FSF 14.2.1 release - it may just be like Alire,
> > which only handles 3 levels, so they call the first packaging of 14.2.0
> > 14.2.1. If you say 'gcc -v' it'll probably say 14.2.0.

$ LANG=en gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-redhat-linux/14/lto-wrapper
OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa
OFFLOAD_TARGET_DEFAULT=1
Target: x86_64-redhat-linux
Configured with: ../configure --enable-bootstrap --enable-languages=c,c++,fortran,objc,obj-c++,ada,go,d,m2,lto --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --with-bugurl=http://bugzilla.redhat.com/bugzilla --enable-shared --enable-threads=posix --enable-checking=release --enable-multilib --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-gnu-unique-object --enable-linker-build-id --with-gcc-major-version-only --enable-libstdcxx-backtrace --with-libstdcxx-zoneinfo=/usr/share/zoneinfo --with-linker-hash-style=gnu --enable-plugin --enable-initfini-array --with-isl=/builddir/build/BUILD/gcc-14.2.1-build/gcc-14.2.1-20240912/obj-x86_64-redhat-linux/isl-install --enable-offload-targets=nvptx-none,amdgcn-amdhsa --enable-offload-defaulted --without-cuda-driver --enable-gnu-indirect-function --enable-cet --with-tune=generic --with-arch_32=i686 --build=x86_64-redhat-linux --with-build-config=bootstrap-lto --enable-link-serialization=1
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.2.1 20240912 (Red Hat 14.2.1-3) (GCC) 

> I see some potential for confusion, since there almost certainly will be
> an official gcc 14.2.1 release in the near future.  That official
> release will include code that's not included in what Fedora calls gcc
> 14.2.1.  It's a point release, so I wouldn't expect substantial changes,
> but still, I think Fedora should use a different naming scheme.

Anyone who needs to know the exact state of the code packaged in Fedora
41 can find the Git revision and all the patches here:
https://src.fedoraproject.org/rpms/gcc/blob/f41/f/gcc.spec

If someone has a problem with the version numbering, they could always
try reporting it as a bug, but they should first know that the main
admin of the Fedora package is also frequently seen in the revision
history at gcc.gnu.org. One who is that deeply involved in the
development of GCC probably has a better idea of the differences
between versions than most of us do.

Björn Persson

Re: [ANN] GCC 14.2.0-3 (aarch64, macOS)

[Keith Thompson]

Keith Thompson <Keith.S.Thompson+u@gmail.com> writes:
> Simon Wright <simon@pushface.org> writes:
>> Björn Persson <Bjorn@rombobjörn.se> writes:
>>> It happened today in GCC 14.2.1 (as packaged in Fedora 41), so no, not
>>> only in 14.2.0.
>>
>> There's no official FSF 14.2.1 release - it may just be like Alire,
>> which only handles 3 levels, so they call the first packaging of 14.2.0
>> 14.2.1. If you say 'gcc -v' it'll probably say 14.2.0.
>>
>> Of course I could be completely wrong and Fedora have added lots of
>> value!
>
> I see some potential for confusion, since there almost certainly will be
> an official gcc 14.2.1 release in the near future.  That official
> release will include code that's not included in what Fedora calls gcc
> 14.2.1.  It's a point release, so I wouldn't expect substantial changes,
> but still, I think Fedora should use a different naming scheme.

After I wrote the above, I built gcc from its git repo, using the tip
of the releases/gcc-14 branch (not a release tag).  Presumably the
Fedora folks did something similar.  The resulting gcc reports its
own version as :

    $ gcc --version
    gcc (GCC) 14.2.1 20241129

whereas a gcc built from the releases/gcc-14.2.0 tag reports :

    $ gcc --version
    gcc (GCC) 14.2.0

with no date.

I still see some potential for confusion, but not as much as I initially
thought.  (The fact that gcc doesn't do *.*.1 releases is fairly
obscure, and I wouldn't expect most people to know about it.)

-- 
Keith Thompson (The_Other_Keith) Keith.S.Thompson+u@gmail.com
void Void(void) { Void(); } /* The recursive call of the void */