From: Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
To: Richard Biener <richard.guenther@gmail.com>
Cc: gcc@gcc.gnu.org
Subject: Re: How to debug while using LTO?
Date: Wed, 30 Nov 2022 11:49:13 +0100 [thread overview]
Message-ID: <Y4c1KbrxcnNagQ1E@li-42a4824c-28a0-11b2-a85c-f55c0d5956ce.ibm.com> (raw)
In-Reply-To: <689A7218-A12B-4612-B8C0-EACE12C4F8AC@gmail.com>
On Thu, Nov 24, 2022 at 05:53:53PM +0100, Richard Biener wrote:
>
>
> > Am 24.11.2022 um 17:28 schrieb Stefan Schulze Frielinghaus via Gcc <gcc@gcc.gnu.org>:
> >
> > Hi everyone,
> >
> > Currently I'm looking into a wrong-code bug and would like to understand
> > a certain optimization done by combine during local transformation.
> > Without LTO I would simply debug cc1 and step through combine. However,
> > with LTO enabled AFAIK I have to debug lto1 instead. In order to get
> > the lto1 command line of interest according to
> > https://gcc.gnu.org/legacy-ml/gcc/2009-11/msg00047.html
> > I have to pass -Wl,-debug to gcc in order to get the command for
> > collect2 to which itself I have to pass -plugin-opt=-debug in order to
> > get the command for lto-wrapper. According to the aforementioned mail I
> > should add option -debug to lto-wrapper, however, it appears to me that
> > option -debug was removed. I gave options -v and -### a chance without
> > luck, i.e., those only print the usual environment variables and
> > afterwards a list of object files like
> >
> > /tmp/ccPEIV35.ltrans0.ltrans.o
> > /tmp/ccNmpKfS.debug.temp.o
> > /tmp/cceiCIFg.debug.temp.o
> > /tmp/ccZ4Qc7E.debug.temp.o
> > ...
> >
> > but no lto1 command. Thus, how do you retrieve the lto1 command?
> >
> > While desperate I retrieved it manually via strace. However, the lto1
> > command refers to temporary files which have been erased meanwhile. I
> > actually didn't expect that because I added -save-temps to all the
> > intermediate commands which is also reflected in the environment
> > variable COLLECT_GCC_OPTIONS. Thus, how do you keep temporary files?
>
> Adding -v -save-temps and then running gdb on the lto1 command works and is what I usually do.
Strangely this is not the case for the current project I'm looking at. The
link step is done via
/usr/local/bin/s390x-linux-gnu-gcc -save-temps \
-O3 -DNDEBUG -flto=auto -fno-fat-lto-objects \
-march=z15 -static -nostdlib \
-Wl,--fatal-warnings -Wl,--no-warn-rwx-segments \
-Werror -Wall -Wundef -T/devel/foo.ld \
obj1.o ... objN.o
where -save-temps is given and was also given during compile time of each
object obj1.o ... objN.o file. Anyhow, adding -Wl,-debug to the link command
above I see that -save-temps is also included in the environment variable
COLLECT_GCC_OPTIONS="... -save-temps ..." and I get the linker command
/usr/local/lib/gcc/s390x-linux-gnu/13.0.0/../../../../s390x-linux-gnu/bin/ld \
-plugin /usr/local/libexec/gcc/s390x-linux-gnu/13.0.0/liblto_plugin.so \
-plugin-opt=/usr/local/libexec/gcc/s390x-linux-gnu/13.0.0/lto-wrapper \
-plugin-opt=-fresolution=foo.res \
-m elf64_s390 -static -o foo \
-L/usr/local/lib/gcc/s390x-linux-gnu/13.0.0 \
-L/usr/local/lib/gcc/s390x-linux-gnu/13.0.0/../../../../s390x-linux-gnu/lib \
--fatal-warnings --no-warn-rwx-segments \
obj1.o ... objN.o -lgcc -T /devel/foo.ld
to which I add -plugin-opt=-debug in order to get the lto-wrapper command
/usr/local/libexec/gcc/s390x-linux-gnu/13.0.0/lto-wrapper \
-fresolution=foo.res -flinker-output=exec \
obj1.o ... objN.o
to which I add -v -save-temps in order to actually get the lto1 command. From
the output printed to stderr I see one lto1 command:
/usr/local/libexec/gcc/s390x-linux-gnu/13.0.0/lto1 -quiet -dumpbase ./a.wpa \
-march=z15 -m64 -mzarch -g -O3 -version -fno-openmp -fno-openacc -fno-pie \
-fcf-protection=none -fltrans-output-list=/tmp/ccsuw4Nj.ltrans.out \
-fwpa=32 -fresolution=foo.res -flinker-output=exec @./a.wpa.args.0
However, this lto1 invocation does no local transformation. Having a look at
the strace output I see a second lto1 invocation:
/usr/local/libexec/gcc/s390x-linux-gnu/13.0.0/lto1 -quiet \
-dumpbase ./a.ltrans0.ltrans -march=z15 -m64 -mzarch -g -O3 \
-fno-openmp -fno-openacc -fno-pie -fcf-protection=none \
-fltrans /tmp/ccsuw4Nj.ltrans0.o -o /tmp/cc69wMp3.s
This lto1 invocation seems to originate from Makefile /tmp/cc5piuXT.mk
generated by lto-wrapper. To my surprise the Makefile as well as the temporary
file /tmp/ccsuw4Nj.ltrans0.o are removed although -save-temps was given all the
time. Stepping through lto-wrapper with gdb I see that
save_temps = 1;
is never executed while walking over decoded_options.
For the time being I hard coded save_temps=1 into lto-wrapper.cc and I am now
able to debug the lto1 invocation of interest. Not sure whether there
is a more elegant way?
All this is based upon a rather huge CMake based project and so far I didn't
find the time to come up with a small reproducer. Still hoping this might be
useful for someone else.
Cheers,
Stefan
next prev parent reply other threads:[~2022-11-30 10:49 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-24 16:26 Stefan Schulze Frielinghaus
2022-11-24 16:53 ` Richard Biener
2022-11-30 10:49 ` Stefan Schulze Frielinghaus [this message]
2022-11-30 13:22 ` Richard Biener
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Y4c1KbrxcnNagQ1E@li-42a4824c-28a0-11b2-a85c-f55c0d5956ce.ibm.com \
--to=stefansf@linux.ibm.com \
--cc=gcc@gcc.gnu.org \
--cc=richard.guenther@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).