On Wed, Oct 23, 2002 at 05:29:40PM -0400, Andrew Gallatin wrote: > > > > Ruslan Ermilov writes: > > OK, to summarize things. There was a single problem with two > > symptoms: 1) groff, if built dynamically, could not be run > > by ld-elf.so; 2) groff, if built statically, always failed > > with ``out of memory'', apparently due to the same bug. > > > > Static hack is safe to delete because: > > > > 1. groff that is built as part of the bootstrap-tools during > > buildworld will be built static anyway (see -DNOSHARED in > > BMAKE in Makefile.inc1) > > > > 2. if you have a vulnerable kernel and rtld-elf, static > > linkage does not address the problem -- you get spurious > > ``out of memory'' even if you link groff statically. > > > > If you agree, please feel free to commit the backout of > > the hack yourself -- I'm going to leave the computer now. :-) > > Removed. > > Please review the following UPDATING entry: > > --- UPDATING 3 Sep 2002 06:13:43 -0000 1.217 > +++ UPDATING 23 Oct 2002 21:24:44 -0000 > @@ -22,6 +22,19 @@ > integrity. Re-enabling write caching can substantially > improve > performance. > > +20021023: > + Alphas with kernels from between 20020902 and 20021022 and/or > + rtld (ld-elf.so.1) older than 20021022 may experience problems > + with groff while doing a buildworld (kernel: "out of memory", > + rtld: "too few PT_LOAD segments"). > + > + So, to successfully upgrade your Alpha, you must either > + upgrade your kernel and rtld first (which might be a bit > + tricky), or avoid running the bootstrapped groff during the > + "transitional" buildworld. To avoid running groff during the > + transitional upgrade run make buildworld with -DNOMAN, > + -DNO_SHAREDOCS, and -DNO_LPR. > + > 20020831: > gcc has been upgraded to 3.2. It is not all binary compatible > with earlier versions of gcc for c++ programs. All c++ > > > Note: I have NOT tested this, beyond verifying that a kernel from Sep > 02 works fine. > What commit is responsible for a breakage, this one?
: peter 2002/09/03 14:18:17 PDT : : Modified files: : sys/kern imgact_elf.c : Log: : Make the text segment locating heuristics from rev 1.121 more reliable : so that it works on the Alpha. This defines the segment that the entry : point exists in as 'text' and any others (usually one) as data. : : Submitted by: tmm : Tested on: i386, alpha : : Revision Changes Path : 1.125 +10 -15 src/sys/kern/imgact_elf.c Also I have verified (on beast) that previous version of Groff (1.17.1) does not have this problem -- groff has two PT_LOAD segments. Anyone cares to explain what's going on here? Why when I remove -fno-exceptions it creates two PT_LOAD segments, why it does not affect previous version of Groff? Cheers, -- Ruslan Ermilov Sysadmin and DBA, [EMAIL PROTECTED] Sunbay Software AG, [EMAIL PROTECTED] FreeBSD committer, +380.652.512.251 Simferopol, Ukraine http://www.FreeBSD.org The Power To Serve http://www.oracle.com Enabling The Information Age
msg45130/pgp00000.pgp
Description: PGP signature
