https://sourceware.org/bugzilla/show_bug.cgi?id=32732
Jan Beulich <jbeulich at suse dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Status|UNCONFIRMED |NEW
Last reconfirmed| |2025-03-10
CC| |jbeulich at suse dot com
--- Comment #2 from Jan Beulich <jbeulich at suse dot com> ---
In the course of trying to adapt the testcase that the original patch added, to
fit the (vaguely) change proposed in
https://sourceware.org/pipermail/binutils/2025-March/139831.html
I came to wonder what the original purpose of that change was:
--section-alignment, like several other command line options controlling pe_*
variables, is imo supposed to affect PE binaries only. Object files should be
left alone. This can perhaps be seen even more prominently by the odd side
effect --subsystem has on object files (for certain option arguments) with the
original change in place. That option yet more clearly ought to be entirely
benign to object files (and we may even want to warn about use of such options
on object files).
For executables, however, playing with section alignment is pointless, as
section alignment is deliberately omitted from section table entries there (as
of Alan's 6f8f6017a0c4 ["PR27567, Linking PE files adds alignment section flags
to executables"]).
Furthermore --set-section-alignment is what is (aiui) supposed to be used when
trying to control alignment of sections in an object file. (I have a patch to
enable the binutils-all/set-section-alignment test for COFF as well, but I
first need to see how well it fares over a fuller range of targets.)
Therefore my question is: Shouldn't we revert the non-auxiliary part of that
change (and even backport that to at least 2.44)? Or at least, extending the
patch sent to Bastian, also drop the inner if() from the conditional in
question?
Orthogonal to the issue here, but an observation while looking at that change
in detail: In the VMA and LMA checks added, why is it "alignment !=
bfd_section_alignment (isection)" rather than "alignment >
bfd_section_alignment (isection)"? Reducing alignment, after all, isn't going
to break a section that was already misaligned.
--
You are receiving this mail because:
You are on the CC list for the bug.
