Re: building ICU 2.6 on Irix 6.5

[George R. <gr...@us...>]

As the readme.html that comes with ICU states, the Irix and Tru64 
platforms are rarely tested.  You will have to port ICU to Irix again or 
finish the Tru64 port.

Test failures can mean one of several things.

1) Your changes to gennorm don't work.
2) The compiler optimized something incorrectly.
3) The config/mh-irix file is out of date or doesn't work with your 
configuration.
4) The configure script detected something wrong, and wrote the 
common/unicode/platform.h file incorrectly.  (e.g. int32_t is the wrong 
size, which can happen if it's typedef'd to a long and it's a 64-bit 
compiler and not a 32-bit compiler)
5) It's a lurking bug in our code that somehow managed to work on many 
other platforms.

A patch file or a more detailed description of your changes would also 
help to diagnose the problem.

George Rhoten
IBM Globalization Center of Competency/ICU  San José, CA, USA
ICU main website: http://oss.software.ibm.com/icu/index.html




"Bradshaw, Karen" <Kar...@io...>
Sent by: icu...@ww...
01/30/2004 07:20 AM
 
        To:     <icu...@ww...>
        cc: 
        Subject:        building ICU 2.6 on Irix 6.5




Have built ICU 2.6 on Irix6.5, MIPSpro version 7.4.  Ran into a number of 
problems in building the data library.  I ran the test suite and found 
many errors in collation and normalization (C++ tests).  The genorm tool 
core'd during the build of the data library.  I had to patch this code to 
proceed.  Other files required changes such as uperf.c(time.h vs. 
sys/time.h) and tsmthred.cpp(use of sleep/usleep errors).  By the results 
of the testsuite, should I assume that the gennorm tool is not 
functioning?

BTW, I had little problems building ICU 2.6 on Tru64.  I used Gmake and 
updated pkgdata tool to do the same.

rgds,

Karen Bradshaw
IONA Technologies