Re: [icu-support] Building ICU on z/OS

[Tom Q. <tom...@te...>]

I’ve essentially tried that – I in fact used the online data configurator to slim down the data package because I wondered whether pkgdata was objecting to the amount of data or something.

If I run gmake in the source/data directory, the tail of what I get is:



pkgdata:  cc  -DOS390BATCH -D_OPEN_THREADS -2 -Wc,'inline(auto,noreport,500,4000)'  -Wc,'langlvl(extended),ros,rent,dll' -DIEEE_754=0 -Wc,'xplink(backchain,storeargs),goff' -c -I../common -I../common -D_SHR_TZNAME -D_SHR_TIMEZONE -Wc,expo -o ./out/tmp/windows_949_2000_cnv.o ./out/tmp/windows_949_2000_cnv.c

pkgdata:  cc  -DOS390BATCH -D_OPEN_THREADS -2 -Wc,'inline(auto,noreport,500,4000)'  -Wc,'langlvl(extended),ros,rent,dll' -DIEEE_754=0 -Wc,'xplink(backchain,storeargs),goff' -c -I../common -I../common -D_SHR_TZNAME -D_SHR_TIMEZONE -Wc,expo -o ./out/tmp/windows_950_2000_cnv.o ./out/tmp/windows_950_2000_cnv.c

pkgdata:  cc -2 -Wc,'inline(auto,noreport,500,4000)'  -Wc,'langlvl(extended),ros,rent,dll' -DIEEE_754=0 -Wl,dll  -Wl,xplink  -o ../lib/libicudata44.1.dll  ./out/tmp/icudt44e_dat.o ./out/tmp/cns_11643_1992_cnv.o ./out/tmp/cnvalias_icu.o ./out/tmp/confusables_cfu.o ./out/tmp/ebcdic_xml_us_cnv.o ./out/tmp/gb18030_cnv.o ./

....

...

...

/out/tmp/windows_874_2000_cnv.o ./out/tmp/windows_936_2000_cnv.o ./out/tmp/windows_949_2000_cnv.o ./out/tmp/windows_950_2000_cnv.o

# Creating data archive library file ..

gmake: *** [packagedata] Killed







From: Steven R. Loomis [mailto:sr...@ic...]
Sent: 30 September 2010 20:18
To: icu...@li...
Subject: Re: [icu-support] Building ICU on z/OS



If you have built the same ICU on another system (even a different platform), copy the dat file from source/data/out/tmp/icudt45l.dat or source/data/out/tmp/icudt45b.dat on that system  into the source/data/in/  directory on Z/OS. That will let the build start with pre-built data.  Then try make in source/data.

-s

On 9/30/10 11:06 AM, Tom Quarendon wrote:

OK thanks. I’ll check out the updated FAQ.

So are you saying that with 4.4.1 it won’t work – I’ll just need to wait for 4.4.2?



What kind of details would you be after for the killed process?

Literally all I get is “Killed”. You get output from pkgdata indicating each of the files that it has compiled, and it appears to have done them all (if I turn on the verbose option to pkgdata, the last entry you get is

# Creating data archive library file ..

After which you get “Killed”.

So it appears to have built everything and is cleaning up when it falls over in some way.

I’ve never really understood what getting just “Killed” from a process on z/OS means.



Tom.



From: Steven R. Loomis [mailto:sr...@ic...]
Sent: 30 September 2010 18:43
To: icu...@li...<mailto:icu...@li...>
Subject: Re: [icu-support] Building ICU on z/OS



Hello,
 I've updated the FAQ about this data member:
    <http://userguide.icu-project.org/icufaq?pli=1#TOC-Why-am-I-seeing-a-small-only-a-few-><http://userguide.icu-project.org/icufaq?pli=1#TOC-Why-am-I-seeing-a-small-only-a-few->

pkgdata is what builds the IXMI44DA file.

However, you shouldn't need to build anything from icudata.lst - ICU 4.4.1 was not packaged properly. ICU 4.4.2 will be coming out shortly (this week), and should build with more success on Z.

Can you get more details on the 'killed' process?

Steven

On 9/30/10 7:47 AM, Tom Quarendon wrote:

I’m having a couple of issues building ICU on z/OS that I was hoping someone could help me with.

The first occurs in the pkgdata phase, whereby the pkgdata process gets “Killed”. All processing _appears_ to have been performed, in that the last entry in the icu/source/data/out/tmp/icudata.lst file appears to have been processed, but then you just get a

“Killed”

message on the shell console and gmake stops. I ignored the return code from the pkgdata command (put ‘-‘ in front of it in the make file), and it appears to have built, but it’s not obvious what the issue is, or whether indeed it has built correctly. May be related to the next issue.



The second issue is that I’m attempting to build to z/OS data sets as well as USS. All appears to get built OK, and I end up with a PDSE with IXMI44IN, IXMI44IO, IXMI44LE, IXMI44LX, IXMI44UC and IXMI44DA in it, which presumably correspond to the normal ICU library names. The issue is that the data file (IXMI44DA) is small, only 0x580 bytes long, and if I try and run a program that loads a converter I get a null pointer back. So this clearly has no actual data in it.

How do I build a “proper” IXMI44DA member? Am I supposed to build a “proper” IXMI44DA member, or is it done some other way?



Any help on this (especially the second) would be great.

Thanks.









--
Steven R. Loomis  sr...@ic...<mailto:sr...@ic...>
Technical Lead, ICU for C/C++   <http://icu-project.org><http://icu-project.org>
IBM San José Globalization Center of Competency
<http://ibm.com/software/globalization><http://ibm.com/software/globalization>



__________ Information from ESET NOD32 Antivirus, version of virus signature database 5493 (20100930) __________

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com