https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119977
Sam James <sjames at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #61222|0 |1
is obsolete| |
--- Comment #11 from Sam James <sjames at gcc dot gnu.org> ---
Created attachment 61229
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=61229&action=edit
bisect.sh
OK, rewinding a bit.
Immediately after the bootstrap failure, without rebuilding anything to get
preprocessed sources (maybe I broke something somehow when doing that), I then
used `compare-lto -p ...` and with `git diff ...` as above, I see:
```
+++ b/dev/fd/62
@@ -1,5 +1,5 @@
LTO Object Name:
[-gimple-match-6-stage2.o.stripped-]{+gimple-match-6-stage3.o.stripped+}
No. Offset Size Section Name
@@ -8,233 +8,233 @@ No. Offset Size Section Name
3 869857 8147 .gnu.lto_.icf.1
4 878004 6179 .gnu.lto_.ipa_sra.1
5 884183 18958 .gnu.lto_.inline.1
6 903141 [-19955-]{+19940+} .gnu.lto_.jmpfuncs.1
7 [-923096-]{+923081+} 559 .gnu.lto_.pureconst.1
8 [-923655 3301-]{+923640 3302+} .gnu.lto_.ipa_modref.1
9 [-926956-]{+926942+} 8 .gnu.lto_.lto.1
10 [-926964-]{+926950+} 549 .gnu.lto__ZN16wide_int_storageC2ERKS_.0.1
11 [-927513-]{+927499+} 840 .gnu.lto__ZN16wide_int_storageaSERKS_.2.1
[...]
```
i.e. the actual size of .gnu.lto_.jmpfuncs.1 and .gnu.lto_.ipa_modref.1 differ
slightly, not just the offsets.
Attaching:
* the stage2 & stage3 objects
* log of the build (in case I'm missing something)
* new script (just s/bootstrap-lean//)
