Hi Jeff,
On 6/12/25 16:14, Jeff Law wrote:
On 11/26/25 7:45 PM, Kalvis Duckmanton wrote:
On 26/11/25 11:15, Jeff Law wrote:
On 11/18/25 7:41 PM, Kalvis Duckmanton wrote:
Hi!
The PCH use_address hooks for NetBSD hosts have not yet been
updated to allow compiled headers to be loaded at an address
different from their preferred address.
This change updates host-netbsd.cc:netbsd_gt_pch_use_address()
thus: if a compiled header cannot be mapped at its preferred
address, a region of memory is allocated and the base address of
this region is passed back to the caller
(ggc-common.cc:gt_pch_restore() I believe). Note that in this case
the return value is 0, allowing gt_pch_restore() to load the
header. In this respect the behaviour is slightly different from
that of the use_address hook for other hosts (e.g. Linux).
This change against GCC 15.2.0 builds on the work in pch/71934 (and
target/58937)
ChangeLog:
* gcc/config/host-netbsd.cc (netbsd_gt_pch_use_address):
update for pch/71934
Generally it looks sensible. So my preference would be to fix both
NetBSD and OpenBSD at the same time, assuming we think both need the
same fix.
Ok. I will investigate a bit further and try to confirm that the
proposed change will also work for OpenBSD.
It'd be hugely appreciated. I'm an old BSD guy myself and would love
to see it better supported these days, but can't really spend any time
on it. In theory this ought to be testable in a VM.
I ended up netbooting OpenBSD/amd64 7.8 and compiling GCC 15.2.0 from
source. In addition to my patch, I also needed to apply the patches
from lang/gcc/15 in the OpenBSD ports repository (a converted version of
which is at https://github.com/openbsd/ports).
Having updated host-openbsd.cc with the same change as for NetBSD, I
tested the change by running cc1plus under lldb. Using lldb, the first
test in openbsd_gt_pch_use_address() (for addr == base) was skipped, and
the address returned by the second mmap() call was modified to be 16kB
above its original value. Once cc1plus had finished, the assembly code
it produced was inspected. The output seemed reasonable and it
contained values defined in the compiled header, which is encouraging.
I've attached an updated version of the previous patch which updates the
use_address hook for NetBSD and for OpenBSD.
cheers
kalvis
diff -cwr a/gcc/config/host-netbsd.cc b/gcc/config/host-netbsd.cc
*** a/gcc/config/host-netbsd.cc Fri Aug 8 16:51:40 2025
--- b/gcc/config/host-netbsd.cc Thu Nov 13 12:39:29 2025
***************
*** 78,84 ****
addr = mmap (base, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_FIXED, fd, offset);
! return addr == base ? 1 : -1;
}
--- 78,98 ----
addr = mmap (base, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_FIXED, fd, offset);
! if (addr == base)
! return 1;
!
! if (addr != (void *) MAP_FAILED)
! munmap(addr, size);
!
! addr = mmap (base, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
!
! if (addr == (void *) MAP_FAILED)
! return -1;
!
! /* Signal to the caller that whilst memory has been allocated, it
! must read the PCH data */
! base = addr;
! return 0;
}
diff -cwr a/gcc/config/host-openbsd.cc b/gcc/config/host-openbsd.cc
*** a/gcc/config/host-openbsd.cc Fri Aug 8 16:51:40 2025
--- b/gcc/config/host-openbsd.cc Sun Dec 21 14:51:12 2025
***************
*** 78,84 ****
addr = mmap (base, size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, offset);
! return addr == base ? 1 : -1;
}
--- 78,98 ----
addr = mmap (base, size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, offset);
! if (addr == base)
! return 1;
!
! if (addr != (void *) MAP_FAILED)
! munmap(addr, size);
!
! addr = mmap (base, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
!
! if (addr == (void *) MAP_FAILED)
! return -1;
!
! /* Signal to the caller that whilst memory has been allocated, it
! must read the PCH data */
! base = addr;
! return 0;
}
