This is the mail archive of the
mailing list for the binutils project.
Re: [PATCH v2 0/6] endianness fixes and preparation for strtab work
- From: jose dot marchesi at oracle dot com (Jose E. Marchesi)
- To: Nick Alcock <nick dot alcock at oracle dot com>
- Cc: binutils at sourceware dot org, indu dot bhagat at oracle dot com
- Date: Fri, 21 Jun 2019 14:14:06 +0200
- Subject: Re: [PATCH v2 0/6] endianness fixes and preparation for strtab work
- References: <email@example.com>
This collection of roughly-related acccumulated patches fixes a bunch of
bugs related to big-endian usage and cross-endian operation (reading
files on one endianness that were generated on another).
It also fixes a few nits involving lack of diagnostics of various forms
of file corruption and out-of-memory errors and prepares for upcoming
patches by discarding the mmap-based memory allocator: we always use
malloc() now, and no longer try to mprotect() our storage.
The endianness fixes trigger one format change: the struct ctf_slice_t
is naturally aligned now, even though this wastes space: slices are
expected to be relatively rare, and not aligning it causes the entire
rest of the file to be unaligned, necessitating memmove() to access
all future elements -- which we are not doing, so platforms incapable
of unaligned I/O crash instead. This format change has no implications
for existing users, since nothing generates slices yet, so there is no
format version bump and we can keep using CTF_VERSION_3.
The test matrix has expanded to cover opening of foreign-endian
.ctf-containing ELF files and foreign-endian raw CTF and CTFA files on
both big- and little-endian platforms (for as long as I continue to have
access to them, that is: in particular I'm not sure how long I'll have
access to a Windows box for mingw tests, since my ancient Windows 7
laptop is about to be retired).
Changes since version 1:
- cover letter
- The ctf_open endianness patch is new.
Pushed in your behalf.