Re: macOS status

"Kirill A. Korinsky" <[email protected]> Wed, 27 Dec 2023 23:50:26 +0100
Newsgroups	gmane.lisp.gcl.devel
Message-ID	<[email protected]>
Hello,

> On 26. Dec 2023, at 14:33, Camm Maguire <[email protected]> wrote:
> 
> Greetings, and happy holidays!
> 
> I could use a little summary as to the live remaining issues.  My
> understanding is that both hunks of your previous 'memory corruption'
> patch are now gone, one because posix_spawn is in the tree, and the
> second as you report you cannot reproduce the file closing issue.
> Working from the same commit is obviously preferable, but a few
> explicitly mentioned patches are ok too with each report.  Logs or
> pointers to logs are great too!
> 
> We also appear to have several simultaneous environments, private and
> github/macports, x86_64, i386, arm/rosetta (? not sure what this is),
> perhaps different macos versions (12, 13?), gcc vs. clang, etc.
> Probably best to keep a little table and tackle one box at a time.  The
> posted build log at the link you sent is very helpful, as was the
> backtrace of the recent environment variable processing we discovered.
> (This needs a comprehensive fix along the lines of your atof which I
> will work out with you.)

I agree to move on one issue per time to avoid any confusion. Let start from some the issue an out of bounds array access was attempted.

(I do have dozen of virtual machines and a few physical machines which may run arm64, x86_64, i386 and PPC with some limitations, let move to one step on the time).

> The issue below means an out of bounds array access was attempted.  I
> cannot reproduce at the commit you mention.  Almost surely something
> else has been miscompiled here.  May I suggest seeing if you can
> reproduce this with current HEAD, and if so, bisecting from your last
> successful build.  Assuming x86_64/clang as before.


I can reproduce it by building it from console on my laptop (macOS 12) with script:

> #!/usr/bin/env sh
> 
> set -e
> 
> make clean || :
> 
> export CC='/usr/bin/clang'
> export CFLAGS='-pipe -fno-pie -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk -arch x86_64 -Wno-implicit-function-declaration -g -O0'
> export CPATH='/opt/local/include'
> export CPPFLAGS='-I/opt/local/include -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk'
> export CXX='/usr/bin/clang++'
> export CXXFLAGS='-pipe -Os -stdlib=libc++ -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk -arch x86_64 -g -O0'
> export DEVELOPER_DIR='/Library/Developer/CommandLineTools'
> export F90FLAGS='-pipe -Os -m64'
> export FCFLAGS='-pipe -Os -m64'
> export FFLAGS='-pipe -Os -m64'
> export INSTALL='/usr/bin/install -c'
> export LDFLAGS='-L/opt/local/lib -Wl,-headerpad_max_install_names -fno-pie -Wl,-syslibroot,/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk -arch x86_64'
> export LIBRARY_PATH='/opt/local/lib'
> export MACOSX_DEPLOYMENT_TARGET='12.0'
> export OBJC='/usr/bin/clang'
> export OBJCFLAGS='-pipe -Os -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk -arch x86_64'
> export OBJCXX='/usr/bin/clang++'
> export OBJCXXFLAGS='-pipe -Os -stdlib=libc++ -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk -arch x86_64'
> export SDKROOT='/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk'
> 
> ./configure --prefix=/opt/local \
>     --disable-notify \
>     --disable-statsysbfd \
>     --enable-ansi \
>     --enable-custreloc \
>     --enable-emacsdir=/opt/local/share/emacs/site-lisp/gcl \
>     --enable-infodir=/opt/local/share/info \
>     --enable-readline \
>     --disable-gprof \
>     --without-x \
>     --disable-xgcl \
>     --host=x86_64-apple-darwin \
>     --build=x86_64-apple-darwin \
>     --target=x86_64-apple-darwin
> 
> make


It setups a bit more ENV variables that nessesary but it emulates behaviour of MacPorts.

So, I confirm that crash on local root 4bf00f6bb804fae2b74366346a52482f740a01e6 and before run this script I've clean everything by: git clean -xdf; git reset --hard

The error was exactly the same, and was triggered and the same place.

> 
> "Kirill A. Korinsky" <[email protected]> writes:
> 
>> I've localized the statment from boot.lisp which cause (2)
>> 
>> COMPILER>(mapc 'compile (nconc #-pre-gcl '(sbit si::aset si::improper-consp mapcar mapcan mapc mapl
>>                                         ;maplist member member-if member-if-not
>>                                         ;assoc assoc-if assoc-if-not
>>                                         ;rassoc rassoc-if rassoc-if-not
>>                                        )
>>                       'si::(listp ibb ib typep <= coerce < > >= + - set-array 0-byte-array-self set-0-byte-array-self concatenate eql-is-eq)
>>                       '(info-p info-ref info-type info-flags info-ch info-ref-ccb info-ref-clb c1constant-value-object
>>                              var-p var-name var-flags var-kind var-ref var-ref-ccb var-loc var-dt var-type var-mt var-tag var-store
>>                              c-array-rank c-array-dim c-array-elttype c-array-self c-array-hasfillp
>>                              array-dimension array-row-major-index row-major-aref si::row-major-aset
>>                              si::row-major-aref-int aref array-rank array-total-size
>>                              array-has-fill-pointer-p length memoized-hash-equal)))
>> 
>> Correctable error: The assertion (APPLY (QUOTE ARRAY-IN-BOUNDS-P) ARRAY ...) failed.
>> Signalled by APPLY.
>> If continued: Repeat assertion.
>> The assertion (APPLY (QUOTE ARRAY-IN-BOUNDS-P) ARRAY ...) failed.
>> 
>> Broken at LET*.  Type :H for Help.
>>     1 (continue) Repeat assertion.
>>     2  Return to top level.
>> COMPILER>>
> 

--
wbr, Kirill
signature.asc (application/pgp-signature, 833 B)