On 30-Jul-99 Peter Putzer wrote: > On Fri, 30 Jul 1999, Alex Zepeda wrote: > >> On Fri, 30 Jul 1999, Peter Putzer wrote: >> >> > On Fri, 30 Jul 1999, Alex Zepeda wrote: >> > >> > > > > ** can this be done for kde-1.1.2 ? kdvi has become **UNUSABLE** >> > > > > with RedHat 6.0 ..... (and maybe other glibc-2.1.1 systems (?) ) >> > > >> > > It seems to me someone should unbreak glibc instead. What exactly needs >> > > fixing? >> > >> > Well, it looks like a fix for kdvi is more practical, doesn't it? >> >> What? And perpetuate broken software on the masses? >> >> Hmm. Damn this calls for a snyde comment about RMS and the FSF, but I'm >> too tired to make one. > > We don't KNOW what exactly is broken, maybe the autoconf-tests are wrong? > I wouldn't be so quick to put the blame on glibc (but of course it may > indeed be at fault...) > > Peter kdvi uses a kpathsea routine xputenv() for writing environment variables many times. When it was compiled, xputenv() needed to know whether the system's putenv() call is "smart" enought to use malloc() or not. A (kpathsea) autoconf test is run to see whether putenv() is "smart". On redhat-5.2 (glibc-2.0) the test says "yes", xputenv.c is compiled with SMART_PUTENV defined, and the kdvi binary works (on both RedHat 5.2 and 6.0) On redhat-6.0, (glibc-2.1.1) the test says "no", and xputenv.c is compiled with SMART_PUTENV not defined, and kdvi is broken. Overriding the autoconf test with a #define SMART_PUTENV 1 patched into xputenv.c *seems* to fix the problem..... Its hard to believe that the upgrade of glibc from 2.0 to 2.1.1 would involve putenv() becoming "not smart", so I would guess that some change in glibc may have invalidated the test that kpathsea's autoconf script uses (?) duncan ---------------------------------- E-Mail: Duncan Haldane Date: 30-Jul-99 Time: 10:39:28 This message was sent by XFMail ----------------------------------