The newly built shared libraries in the gcc directory aren't always used for bootstap and check, which may lead to the bootstrap failure and false alarm in "make check".
When LD_LIBRARY_PATH is set, dejagnu tests can't use the global ld_library_path since it will cause LD_LIBRARY_PATH to be set with $ld_library_path, which will override the existing LD_LIBRARY_PATH.
*** Bug 17087 has been marked as a duplicate of this bug. ***
Confirmed.
Created attachment 7300 [details] log of failing bootstrap on ia64-unknown-linux-gnu
I observe something similar when trying to boot the trunk on ia64-*-linux-gnu. Specifically, during stage 2, the compiler builds build/genmodes and then runs it, but there is a failure because the just-built libunwind.so.7 is not found: it is in the stage1 subdirectory which is not on any LD_LIBRARY_PATH. I have attached a log from a failing build as an attachment.
It is a known bug. See bug 17783.
Really the libunwind patch should be reverted untill we fix this.
Shouldn't we also revert libgcc_s.so patch? The problem is nothing new. It existed from day one when libgcc_s.so was introduced. If there is no system libgcc_s.so, gcc won't boostrap nor pass "make check". Let's fix this long standing bug.
This is not true at all because I used to bootstrap on linux right after gcc added libgcc_s, In fact I still bootstrap on targets where libgcc_s don't exist on the machine but is built now on the mainline (ppc- darwin is the one which I am talking about).
Have you looked at http://gcc.gnu.org/ml/gcc/2004-09/msg00209.html We get lucky that only Ada compiler is linked with libgcc_s.so.
All the bugs upon which this meta-bug depends have been closed, so I'm going to close this one, too. Feel free to reopen if more bugs crop up that need to be grouped under this one.