[illumos-Developer] Review 693
Igor Kozhukhov
ikozhukhov at gmail.com
Wed Jul 13 13:20:51 PDT 2011
Hi Gordon,
Thanks for your comments - will fix.
About CPPFLAGS and $(ROOT)/usr/include for some packages.
I have removed installed headers system/library/math/header-math from system
and some packages couldn't found installed headers to $(ROOT)/usr/include.
I think it is mistakes in packages and need investigate it and fix it - I'll
try.
About $(SUBDIRS) - i386 and amd64 subdirs have different list of libs - i386
have one libmvec_hwcap1.so.1. I couldn't found solution how to provide
different list of targets to different subdirs. I have more the one lib.
I'll try to investigate how to do it better.
Maybe you can provide me example - maybe package where can I learn better
integration.
At this moment I have used LIBC as template and saved current LIBM
structure.
You can see all files - intel, sparc - I didn't remove some sources.
I have modified only Makefiles and resolved small compilation issues in
sources.
I don't have SPARC environment for testing and resolving compilation issues.
About manifests - I forgot added to HG repo before review.
I'll add to next review, thanks.
This integration have 4 manifests:
usr/src/pkg/manifests/SUNWlibm.mf
usr/src/pkg/manifests/SUNWlibms.mf
usr/src/pkg/manifests/system-library-math-header-math.mf
usr/src/pkg/manifests/system-library-math.mf
SUNWlibm.mf = sysytem-library-math-header-math.mf
SUNWlibms.mf = system-library-math.mf
A have found this example of comparison on OpenSolaris repo.
Best regards,
Igor
> From: Gordon Ross <gordon.w.ross at gmail.com>
> Date: Wed, 13 Jul 2011 15:48:57 -0400
> To: Igor Kozhukhov <ikozhukhov at gmail.com>
> Cc: "developer at lists.illumos.org" <developer at lists.illumos.org>
> Subject: Re: [illumos-Developer] Review 693
>
> On Wed, Jul 13, 2011 at 7:12 AM, Igor Kozhukhov <ikozhukhov at gmail.com> wrote:
>> Hello All,
>>
>> Please review : integration Sun Devpro Math Library to illumos tree (693)
>>
>> http://igork.vlds.ru/nexenta/cr/illumos-693-webrev/
>
> usr/src/Makefile.lint
> This appears to be alphabetical, so please keep it that way.
>
> usr/src/cmd/lms/Makefile
> This part of the include path should come from $(ENVCPPFLAGS1)
> I you use CPPFLAGS += ... this should come in automatically.
>
> usr/src/lib/libsqlite/Makefile.com
> ditto CPPFLAGS += ...
>
> usr/src/cmd/perl/skel/Makefile
> Ditto ENVCPPFLAGS1
> If you must set INC=... here, you should use the appropriate
> variables from usr/src/Makefile.master, i.e. INC=$(CPPFLAGS)
>
> usr/src/lib/Makefile
> The first insert should be two lines later.
> The second and third should be located to
> keep those sections alphabetical.
>
> usr/src/lib/libm/Makefile
> This is an unusual Makefile, doing the install work here,
> instead of in the subdir (per-arch) Makefiles. Why?
>
> usr/src/lib/libm/amd64/Makefile
> Wow! That's one LONG Makefile.
> This could be shortened using pattern expansions...
>
> So, popping "up a level" for a second, where did all these
> come from? Did you write them? Or do you have diffs
> against some previous version of this code? Were there
> sparc assembler files to match usr/src/lib/libm/src/i386?
>
>
> Also, as Alex Eremin commented, I'd suggest getting rid of the
> "src" directory level for new files matching these patterns:
> usr/src/lib/libm/src/mvec/*
> usr/src/lib/libm/src/C/*
> usr/src/lib/libm/src/LD/*
> usr/src/lib/libm/src/Q/*
> usr/src/lib/libm/src/R/*
> usr/src/lib/libm/src/complex/*
> usr/src/lib/libm/src/m9x/*
>
> and for the following subdirs, I suggest moving: (from, to)
> usr/src/lib/libm/src/i386/amd64/ usr/src/lib/libm/ml/amd64/
> usr/src/lib/libm/src/i386/common/ usr/src/lib/libm/ml/common/
> ("ml" for machine language, like under usr/src/uts)
>
> Are there packaging changes still to do?
> ( usr/src/pkg/manifest/* )
>
> Thanks,
> Gordon
>
>> What have been done:
>> 1. integrated LIBM to illumos build for intel i386 and amd64
>> 2. fixed compilation issues in packages for using headers from integrated
>> LIBM
>> Package system/library/math/header-math
>> with headers have been removed from system and build was using integrated
>> LIBM headers
>>
>> This LIBM have been tested on OI-151 and NCP4.
>>
>> This version of LIBM don't have libmtsk*
>>
More information about the Developer
mailing list