------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca 2005-02-13 02:57 ------- Subject: Re: [4.0 Regression] g++.dg/abi/inline1.C fails on hp
> On Sun, 13 Feb 2005, danglin at gcc dot gnu dot org wrote: > > > Looking at the testsuite results, this regression appears to have been > > introduced between Jul 28 22:04:59 UTC and Jul 29 23:04:16 UTC. > > That's when the test was added - but though the test is new since 3.4, > it's still a regression as it passes when built with 3.4.3. > Unsurprisingly it probably wasn't tested on a platform defining > ASM_OUTPUT_EXTERNAL, since most platforms don't. It is quite possible > that that large patch "C++ PATCH: Cleanup linkage handling" did cause the > problem with the test as well - it would be necessary to test compiling > the testcase with compilers from before and after the change to find out, > given that my other endpoint is that 3.4.3 compiles the test to assembler > without the unwanted .IMPORT - but if so then it's likely a side effect of > improving correctness in the C++ front end. (Though of course if there is > a simple fix within the C++ front end which allows the question of > deferring ASM_OUTPUT_EXTERNAL to be put off to 4.1, that would be > convenient.) The above cleanup broke linkage support on systems without weak or one-only support. There is an AIX PR on this. I have access to an old version of the HP linker code which was placed into the public domain. This has been useful in a number of cases in diagnosing problems. I do test HP-UX 10.20 and 11.00, but not on a daily basis. As far as I know, there aren't any significant problems with 4.0.0 beyond that present in 11.11. I currently have PHSS_30968 on the HP-UX 11.11 system that I test on. A linker patch is needed on 11.X to successfully build libstdc++ but I assume you are aware of that. I'll give your change a try when my current build completes. Dave -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19797
