Re: update: math/libcerf and math/gnuplot

2020-03-24 Thread Jeremie Courreges-Anglas 
On Sat, Mar 14 2020, Jeremie Courreges-Anglas  wrote:
> On Fri, Mar 13 2020, Paul Irofti  wrote:
>> Hi,
>>
>> Thank you for your effort! I have a few comments and questions.
>
> I got curious about this proposal, mostly because of the COMPILER line
> change.
>
>> I see they switched to Gitlab (hence the url changes) and that the
>> author set it up for continous integration and thus made it compile with
>> clang.
>>
>> Why did you bump the major. I tried browsing the history on Gitlab to
>> see what changed in the API, but couldn't see much. I am sure I missed
>> something.
>
> As shown by /usr/src/lib/check_sym ("No dynamic export changes"), and
> a diff of the installed include file, the shared lib version bump
> appears unnecessary.
>
>> Is the gnuplot bump necessary? Why?
>
> Maybe Yozo found it useful for testing, but it doesn't seem useful for
> the ports tree.
>
> The updated diff below has been stripped from the items that
> I mentioned.  It also updates WANTLIB, no need for COMPILER_LIBCXX since
> this is a C-only port.  Builds fine on sparc64 along with
> gnuplot,no_x11.  Paul, feel free to use the diff below.

Committed, thanks.

-- 
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF  DDCC 0DFA 74AE 1524 E7EE

Re: update: math/libcerf and math/gnuplot

2020-03-14 Thread Stuart Henderson 
On 2020/03/14 16:55, Yozo TODA wrote:
> yes yes (-:
> that's what I want someone to check if ok, I'm not sure if it's ok
> for gcc compiled application to dynamically load clang-compiled library...

It is normally OK, as long as the C++ standard library is not involved.

Re: update: math/libcerf and math/gnuplot

2020-03-14 Thread Yozo TODA 
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512

> > Why did you bump the major. I tried browsing the history on Gitlab to
> > see what changed in the API, but couldn't see much. I am sure I missed
> > something.

ah, just because I thought it should be bumped when the compiler changed.
no intension other than that.


> As shown by /usr/src/lib/check_sym ("No dynamic export changes"), and
> a diff of the installed include file, the shared lib version bump
> appears unnecessary.

right. I confirmed check_sym too with my local packages. 

  % ls -al /tmp/yozo/*/lib/*.so.*
  -rw-r--r--  1 yozo  wheel  72472 Jan  1  1970 
/tmp/yozo/1.11p0/lib/libcerf.so.2.0
  -rw-r--r--  1 yozo  wheel  69400 Jan  1  1970 
/tmp/yozo/1.13/lib/libcerf.so.3.0
  % /usr/src/lib/check_sym /tmp/yozo/*/lib/*.so.*
  /tmp/yozo/1.11p0/lib/libcerf.so.2.0 --> /tmp/yozo/1.13/lib/libcerf.so.3.0
  No dynamic export changes
  External reference changes:
  added:
floor
  
  removed:
sqrt
  
  PLT removed:
dvoigt
  
  % 

only external dependency diff,
libcerf compiled with gcc depends on sqrt in libm, and
libcerf compiled with clang depends on floor in libm.


> > Is the gnuplot bump necessary? Why?
>
> Maybe Yozo found it useful for testing, but it doesn't seem useful for
> the ports tree.

yes yes (-:
that's what I want someone to check if ok, I'm not sure if it's ok
for gcc compiled application to dynamically load clang-compiled library...

  -- yozo.

-BEGIN PGP SIGNATURE-
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=SSAg
-END PGP SIGNATURE-

Re: update: math/libcerf and math/gnuplot

2020-03-13 Thread Jeremie Courreges-Anglas 
On Fri, Mar 13 2020, Paul Irofti  wrote:
> Hi,
>
> Thank you for your effort! I have a few comments and questions.

I got curious about this proposal, mostly because of the COMPILER line
change.

> I see they switched to Gitlab (hence the url changes) and that the
> author set it up for continous integration and thus made it compile with
> clang.
>
> Why did you bump the major. I tried browsing the history on Gitlab to
> see what changed in the API, but couldn't see much. I am sure I missed
> something.

As shown by /usr/src/lib/check_sym ("No dynamic export changes"), and
a diff of the installed include file, the shared lib version bump
appears unnecessary.

> Is the gnuplot bump necessary? Why?

Maybe Yozo found it useful for testing, but it doesn't seem useful for
the ports tree.

The updated diff below has been stripped from the items that
I mentioned.  It also updates WANTLIB, no need for COMPILER_LIBCXX since
this is a C-only port.  Builds fine on sparc64 along with
gnuplot,no_x11.  Paul, feel free to use the diff below.


Index: Makefile
===
RCS file: /cvs/ports/math/libcerf/Makefile,v
retrieving revision 1.5
diff -u -p -r1.5 Makefile
--- Makefile12 Jul 2019 20:47:42 -  1.5
+++ Makefile14 Mar 2020 00:23:15 -
@@ -2,26 +2,24 @@
 
 COMMENT =  implementation of complex error functions
 
-V =1.11
-DISTNAME = libcerf-${V}
+V =1.13
+DISTNAME = libcerf-v${V}
+PKGNAME =  libcerf-${V}
 EXTRACT_SUFX = .tgz
 CATEGORIES =   math
-MASTER_SITES = http://apps.jcns.fz-juelich.de/src/libcerf/
-REVISION = 0
+MASTER_SITES = https://jugit.fz-juelich.de/mlz/libcerf/-/archive/v${V}/
 
-SHARED_LIBS +=  cerf   2.0 # 1.10
+SHARED_LIBS +=  cerf   2.0 # 1.13
 
-HOMEPAGE = http://apps.jcns.fz-juelich.de/doku/sc/libcerf
+HOMEPAGE = https://jugit.fz-juelich.de/mlz/libcerf/
 
 MAINTAINER =   Paul Irofti 
 
 # MIT
 PERMIT_PACKAGE =   Yes
 
-WANTLIB += c m ${COMPILER_LIBCXX}
+WANTLIB += m
 
 MODULES =  devel/cmake
-
-COMPILER = ports-gcc   # imaginary constants are a GNU extension
 
 .include 
Index: distinfo
===
RCS file: /cvs/ports/math/libcerf/distinfo,v
retrieving revision 1.3
diff -u -p -r1.3 distinfo
--- distinfo29 Dec 2018 07:48:36 -  1.3
+++ distinfo14 Mar 2020 00:23:15 -
@@ -1,2 +1,2 @@
-SHA256 (libcerf-1.11.tgz) = cBAcrEoNeGMyLU0Gz5XFB6nP1k/JmtGzGoQlIEz9lnI=
-SIZE (libcerf-1.11.tgz) = 60066
+SHA256 (libcerf-v1.13.tgz) = 5Gmfga+Diu9bPncgn+yOmCCk9JLVmPtaBwgAhUl2owU=
+SIZE (libcerf-v1.13.tgz) = 60732


-- 
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF  DDCC 0DFA 74AE 1524 E7EE

Re: update: math/libcerf and math/gnuplot

2020-03-13 Thread Paul Irofti 
Hi,

Thank you for your effort! I have a few comments and questions.

I see they switched to Gitlab (hence the url changes) and that the
author set it up for continous integration and thus made it compile with
clang.

Why did you bump the major. I tried browsing the history on Gitlab to
see what changed in the API, but couldn't see much. I am sure I missed
something.

Is the gnuplot bump necessary? Why?

Paul

On Tue, Mar 10, 2020 at 10:44:55AM +0900, Yozo TODA wrote:
> (resending this from gmail, because my ISP's mailserver was blocked
> via UCEProtect Level1. I hope this will go through now...)
> 
> Dear maintainer Paul Irofti, (Cc: ports@openbsd.org)
> here attached the patch to update libcerf and gnuplot;
> please check and commit this if it's OK?
> 
>   - updating libcerf 1.11p0 to 1.13
>   - the upstream site looks renewed (apps.jcns... to jugit...)
>   - I compiled 1.13 successfully with clang, hence removing COMPILER line.
> but I'm not sure it's OK for non-clang architectures...
>   - just adding REVISION=0 to gnuplot to get along with libcerf update,
> and verify that this libcerf update works well.
> all demo work.
> 
> 
> In my environment (following -current, amd64)
> gcc-8 cannot compile for some weeks.
> (some error like "compiler internal error" in the middle of build.)
> So, I cannot rebuild math/maxima. Looking at the build log and Makefile,
> maxima depends on gnuplot depending on libcerf depending on gcc-8.
> That's why I examined libcerf port if it works with clang...
> 
> -- yozo.

> Index: libcerf/Makefile
> ===
> RCS file: /cvs/ports/math/libcerf/Makefile,v
> retrieving revision 1.5
> diff -u -p -r1.5 Makefile
> --- libcerf/Makefile  12 Jul 2019 20:47:42 -  1.5
> +++ libcerf/Makefile  8 Mar 2020 07:08:56 -
> @@ -2,16 +2,16 @@
>  
>  COMMENT =implementation of complex error functions
>  
> -V =  1.11
> -DISTNAME =   libcerf-${V}
> +V =  1.13
> +DISTNAME =   libcerf-v${V}
> +PKGNAME =libcerf-${V}
>  EXTRACT_SUFX =   .tgz
>  CATEGORIES = math
> -MASTER_SITES =   http://apps.jcns.fz-juelich.de/src/libcerf/
> -REVISION =   0
> +MASTER_SITES =   https://jugit.fz-juelich.de/mlz/libcerf/-/archive/v${V}/
>  
> -SHARED_LIBS +=  cerf 2.0 # 1.10
> +SHARED_LIBS +=  cerf 3.0 # 1.13
>  
> -HOMEPAGE =   http://apps.jcns.fz-juelich.de/doku/sc/libcerf
> +HOMEPAGE =   https://jugit.fz-juelich.de/mlz/libcerf/
>  
>  MAINTAINER = Paul Irofti 
>  
> @@ -21,7 +21,5 @@ PERMIT_PACKAGE =Yes
>  WANTLIB += c m ${COMPILER_LIBCXX}
>  
>  MODULES =devel/cmake
> -
> -COMPILER =   ports-gcc   # imaginary constants are a GNU extension
>  
>  .include 
> Index: libcerf/distinfo
> ===
> RCS file: /cvs/ports/math/libcerf/distinfo,v
> retrieving revision 1.3
> diff -u -p -r1.3 distinfo
> --- libcerf/distinfo  29 Dec 2018 07:48:36 -  1.3
> +++ libcerf/distinfo  8 Mar 2020 07:08:56 -
> @@ -1,2 +1,2 @@
> -SHA256 (libcerf-1.11.tgz) = cBAcrEoNeGMyLU0Gz5XFB6nP1k/JmtGzGoQlIEz9lnI=
> -SIZE (libcerf-1.11.tgz) = 60066
> +SHA256 (libcerf-v1.13.tgz) = 5Gmfga+Diu9bPncgn+yOmCCk9JLVmPtaBwgAhUl2owU=
> +SIZE (libcerf-v1.13.tgz) = 60732
> Index: libcerf/patches/patch-man_CMakeLists_txt
> ===
> RCS file: /cvs/ports/math/libcerf/patches/patch-man_CMakeLists_txt,v
> retrieving revision 1.1
> diff -u -p -r1.1 patch-man_CMakeLists_txt
> --- libcerf/patches/patch-man_CMakeLists_txt  28 Dec 2018 16:28:44 -  
> 1.1
> +++ libcerf/patches/patch-man_CMakeLists_txt  8 Mar 2020 07:08:56 -
> @@ -1,4 +1,4 @@
> -$OpenBSD: patch-man_CMakeLists_txt,v 1.1 2018/12/28 16:28:44 pirofti Exp $
> +$OpenBSD$
>  
>  Manual pages should go under ${PREFIX}/man/ rather than under
>  ${PREFIX}/share/man/.
> Index: gnuplot/Makefile
> ===
> RCS file: /cvs/ports/math/gnuplot/Makefile,v
> retrieving revision 1.75
> diff -u -p -r1.75 Makefile
> --- gnuplot/Makefile  8 Nov 2019 23:29:56 -   1.75
> +++ gnuplot/Makefile  8 Mar 2020 07:08:56 -
> @@ -4,6 +4,7 @@ COMMENT = command-driven interactive fun
>  
>  V =  5.2
>  PATCHLEVEL = 7
> +REVISION =   0
>  DISTNAME =   gnuplot-${V}.${PATCHLEVEL}
>  CATEGORIES = math graphics
>  MASTER_SITES =   ${MASTER_SITE_SOURCEFORGE:=gnuplot/}

update: math/libcerf and math/gnuplot

2020-03-09 Thread Yozo TODA 
(resending this from gmail, because my ISP's mailserver was blocked
via UCEProtect Level1. I hope this will go through now...)

Dear maintainer Paul Irofti, (Cc: ports@openbsd.org)
here attached the patch to update libcerf and gnuplot;
please check and commit this if it's OK?

  - updating libcerf 1.11p0 to 1.13
  - the upstream site looks renewed (apps.jcns... to jugit...)
  - I compiled 1.13 successfully with clang, hence removing COMPILER line.
but I'm not sure it's OK for non-clang architectures...
  - just adding REVISION=0 to gnuplot to get along with libcerf update,
and verify that this libcerf update works well.
all demo work.


In my environment (following -current, amd64)
gcc-8 cannot compile for some weeks.
(some error like "compiler internal error" in the middle of build.)
So, I cannot rebuild math/maxima. Looking at the build log and Makefile,
maxima depends on gnuplot depending on libcerf depending on gcc-8.
That's why I examined libcerf port if it works with clang...

-- yozo.
Index: libcerf/Makefile
===
RCS file: /cvs/ports/math/libcerf/Makefile,v
retrieving revision 1.5
diff -u -p -r1.5 Makefile
--- libcerf/Makefile12 Jul 2019 20:47:42 -  1.5
+++ libcerf/Makefile8 Mar 2020 07:08:56 -
@@ -2,16 +2,16 @@
 
 COMMENT =  implementation of complex error functions
 
-V =1.11
-DISTNAME = libcerf-${V}
+V =1.13
+DISTNAME = libcerf-v${V}
+PKGNAME =  libcerf-${V}
 EXTRACT_SUFX = .tgz
 CATEGORIES =   math
-MASTER_SITES = http://apps.jcns.fz-juelich.de/src/libcerf/
-REVISION = 0
+MASTER_SITES = https://jugit.fz-juelich.de/mlz/libcerf/-/archive/v${V}/
 
-SHARED_LIBS +=  cerf   2.0 # 1.10
+SHARED_LIBS +=  cerf   3.0 # 1.13
 
-HOMEPAGE = http://apps.jcns.fz-juelich.de/doku/sc/libcerf
+HOMEPAGE = https://jugit.fz-juelich.de/mlz/libcerf/
 
 MAINTAINER =   Paul Irofti 
 
@@ -21,7 +21,5 @@ PERMIT_PACKAGE =  Yes
 WANTLIB += c m ${COMPILER_LIBCXX}
 
 MODULES =  devel/cmake
-
-COMPILER = ports-gcc   # imaginary constants are a GNU extension
 
 .include 
Index: libcerf/distinfo
===
RCS file: /cvs/ports/math/libcerf/distinfo,v
retrieving revision 1.3
diff -u -p -r1.3 distinfo
--- libcerf/distinfo29 Dec 2018 07:48:36 -  1.3
+++ libcerf/distinfo8 Mar 2020 07:08:56 -
@@ -1,2 +1,2 @@
-SHA256 (libcerf-1.11.tgz) = cBAcrEoNeGMyLU0Gz5XFB6nP1k/JmtGzGoQlIEz9lnI=
-SIZE (libcerf-1.11.tgz) = 60066
+SHA256 (libcerf-v1.13.tgz) = 5Gmfga+Diu9bPncgn+yOmCCk9JLVmPtaBwgAhUl2owU=
+SIZE (libcerf-v1.13.tgz) = 60732
Index: libcerf/patches/patch-man_CMakeLists_txt
===
RCS file: /cvs/ports/math/libcerf/patches/patch-man_CMakeLists_txt,v
retrieving revision 1.1
diff -u -p -r1.1 patch-man_CMakeLists_txt
--- libcerf/patches/patch-man_CMakeLists_txt28 Dec 2018 16:28:44 -  
1.1
+++ libcerf/patches/patch-man_CMakeLists_txt8 Mar 2020 07:08:56 -
@@ -1,4 +1,4 @@
-$OpenBSD: patch-man_CMakeLists_txt,v 1.1 2018/12/28 16:28:44 pirofti Exp $
+$OpenBSD$
 
 Manual pages should go under ${PREFIX}/man/ rather than under
 ${PREFIX}/share/man/.
Index: gnuplot/Makefile
===
RCS file: /cvs/ports/math/gnuplot/Makefile,v
retrieving revision 1.75
diff -u -p -r1.75 Makefile
--- gnuplot/Makefile8 Nov 2019 23:29:56 -   1.75
+++ gnuplot/Makefile8 Mar 2020 07:08:56 -
@@ -4,6 +4,7 @@ COMMENT =   command-driven interactive fun
 
 V =5.2
 PATCHLEVEL =   7
+REVISION = 0
 DISTNAME = gnuplot-${V}.${PATCHLEVEL}
 CATEGORIES =   math graphics
 MASTER_SITES = ${MASTER_SITE_SOURCEFORGE:=gnuplot/}