[core] Reduce symbol search only to when autoloading is enabled. #14287

vgvassilev · 2023-12-20T17:39:50Z

The llvm9 JIT issued callbacks when a symbol was missing and we reacted on it by loading the relevant library. In 9b2041e3 we have kept the logic but now the JIT started querying more often even for symbols which are okay to be missing. In turn that leads to scanning all libraries causing performance issues.

This patch tries to limit this functionality only in contexts where automatic loading is allowed. In addition when computing the offsets of a constant variable declaration we compute the initializers instead of searching in the shared objects.

phsft-bot · 2023-12-20T17:40:00Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

pcanal · 2023-12-20T17:58:31Z

This patch seems to be a good performance enhancement (skipping some work that will not be used) but I am confused on how it helps with the increase in autoloading as I would have expected that in the case where this patch is active (i.e. autoloading is disabled) the net effect would have been the same (i.e. autoloading not activated). What am I missing?

i.e. this patch looks good to me as-is but does not seem to fully address the problem.

github-actions · 2023-12-20T22:56:01Z

Test Results

9 files 9 suites 1d 16h 8m 44s ⏱️
2 488 tests 2 487 ✅ 0 💤 1 ❌
21 360 runs 21 359 ✅ 0 💤 1 ❌

For more details on these failures, see this check.

Results for commit 55145f0.

♻️ This comment has been updated with latest results.

vepadulano · 2023-12-21T11:08:41Z

Reporting some numbers from my machine with this patch, building ROOT with the following cmake invocation:

cmake -DCMAKE_BUILD_TYPE=Release -S rootsrc -B rootbuild/BUILD_NAME -DCMAKE_INSTALL_DIR=rootinstall/INSTALL_NAME

TLDR: The situation w.r.t. #14277 is ~~seemingly unchanged.~~

EDIT: Probably we also want to compare the changes of this patch against ROOT master before the two hotfixes regarding the missing symbols were introduced. I will do that and provide a followup comment.

Baseline: master@fbd6abb

Import time

Taken after a few runs, hot caches

$: time python -c "import ROOT"

real	0m0.791s
user	0m0.592s
sys	0m0.203s

strace file stats

strace -e file python -c 'import ROOT' 2>&1 | awk -v FS='"' '{ print $1 }' | sort | uniq -c
    206 access(
      2 chdir(
      1 execve(
      1 +++ exited with 0 +++
    181 getcwd(
      2 newfstatat(0, 
      2 newfstatat(1, 
      2 newfstatat(2, 
    532 newfstatat(3, 
     29 newfstatat(4, 
  10123 newfstatat(AT_FDCWD, 
    546 openat(AT_FDCWD, 
    843 readlink(
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=130118, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=130120, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=130124, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---

openat stats with the two symbols related to #14277

strace -z -f -o openat.log -e trace=open,openat python3 -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)'
-1
3000
grep openat openat.log | sed 's|.*, "||;s|".*||' | sort | uniq | wc -l
338

After this patch

Import time

Taken after a few runs, hot caches

time python -c "import ROOT"

real	0m0.792s
user	0m0.600s
sys	0m0.197s

strace file stats

strace -e file python -c 'import ROOT' 2>&1 | awk -v FS='"' '{ print $1 }' | sort | uniq -c
    206 access(
      2 chdir(
      1 execve(
      1 +++ exited with 0 +++
    181 getcwd(
      2 newfstatat(0, 
      2 newfstatat(1, 
      2 newfstatat(2, 
    532 newfstatat(3, 
     29 newfstatat(4, 
  10123 newfstatat(AT_FDCWD, 
    546 openat(AT_FDCWD, 
    843 readlink(
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=130864, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=130866, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=130870, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---

openat stats with the two symbols related to #14277

strace -z -f -o openat.log -e trace=open,openat python3 -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)'
-1
3000
grep openat openat.log | sed 's|.*, "||;s|".*||' | sort | uniq | wc -l
338

vgvassilev · 2023-12-21T13:03:23Z

Don’t spend time on this patch just yet. I have more local changes that I am going to upload soon-ish.

phsft-bot · 2023-12-21T21:20:28Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

vgvassilev · 2023-12-21T21:23:09Z

@vepadulano, can you run your tests again. With the current changes we support constexpr and it should be fast as we do not even try to look for the symbols but we evaluate the initializer instead.

phsft-bot · 2023-12-21T21:27:15Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

phsft-bot · 2023-12-21T22:00:20Z

Build failed on ROOT-performance-centos8-multicore/soversion.
Running on olbdw-01.cern.ch:/data/sftnight/workspace/root-pullrequests-build
See console output.

Failing tests:

projectroot.roottest.root.meta.unit.roottest_root_meta_unit_unittests

phsft-bot · 2023-12-21T22:09:17Z

Build failed on ROOT-ubuntu2204/nortcxxmod.
Running on root-ubuntu-2204-2.cern.ch:/home/sftnight/build/workspace/root-pullrequests-build
See console output.

Failing tests:

phsft-bot · 2023-12-21T22:14:34Z

Build failed on mac12arm/cxx20.
Running on 194.12.161.128:/Users/sftnight/build/workspace/root-pullrequests-build
See console output.

Failing tests:

phsft-bot · 2023-12-21T22:19:58Z

Build failed on ROOT-ubuntu2004/python3.
Running on root-ubuntu-2004-1.cern.ch:/home/sftnight/build/workspace/root-pullrequests-build
See console output.

Failing tests:

projectroot.roottest.root.meta.unit.roottest_root_meta_unit_unittests

phsft-bot · 2023-12-21T23:00:39Z

Build failed on windows10/default.
Running on null:C:\build\workspace\root-pullrequests-build
See console output.

Failing tests:

phsft-bot · 2023-12-22T07:24:12Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

phsft-bot · 2023-12-22T07:48:59Z

Build failed on ROOT-ubuntu2204/nortcxxmod.
Running on root-ubuntu-2204-2.cern.ch:/home/sftnight/build/workspace/root-pullrequests-build
See console output.

Failing tests:

projectroot.core.metacling.test.gtest_core_metacling_test_TClingTest

phsft-bot · 2023-12-22T07:53:43Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

phsft-bot · 2023-12-22T08:16:50Z

Build failed on ROOT-ubuntu2204/nortcxxmod.
Running on root-ubuntu-2204-2.cern.ch:/home/sftnight/build/workspace/root-pullrequests-build
See console output.

Failing tests:

projectroot.core.metacling.test.gtest_core_metacling_test_TClingTest

vepadulano · 2023-12-22T08:42:21Z

Updates

Baseline: master before #14261 and #14276

time python -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)'
-1
3000

real	0m1.198s
user	0m0.706s
sys	0m0.475s

nm -C $ROOTSYS/lib/*.so | grep TString::kNPOS | wc -l
0
nm -C $ROOTSYS/lib/*.so | grep kError | wc -l
27

strace -e file python -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)' 2>&1 | awk -v FS='"' '{ print $1 }' | sort | uniq -c
      1 -1
      1 3000
    219 access(
      2 chdir(
      1 execve(
      1 +++ exited with 0 +++
    184 getcwd(
      2 newfstatat(0, 
      2 newfstatat(1, 
      2 newfstatat(2, 
  10455 newfstatat(3, 
   3997 newfstatat(4, 
  31461 newfstatat(AT_FDCWD, 
  19903 openat(AT_FDCWD, 
  28769 readlink(
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=11626, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=11628, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=11632, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---

strace -z -f -o openat.log -e trace=open,openat python3 -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)'
grep openat openat.log | wc -l
19759

Master after the two hotfixes

time python -c "import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)"
-1
3000

real	0m0.807s
user	0m0.620s
sys	0m0.183s

nm -C $ROOTSYS/lib/*.so | grep TString::kNPOS | wc -l
1
nm -C $ROOTSYS/lib/*.so | grep kError | wc -l
34

strace -e file python -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)' 2>&1 | awk -v FS='"' '{ print $1 }' | sort | uniq -c
      1 -1
      1 3000
    219 access(
      2 chdir(
      1 execve(
      1 +++ exited with 0 +++
    184 getcwd(
      2 newfstatat(0, 
      2 newfstatat(1, 
      2 newfstatat(2, 
    532 newfstatat(3, 
     29 newfstatat(4, 
  10141 newfstatat(AT_FDCWD, 
    546 openat(AT_FDCWD, 
    843 readlink(
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12038, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12040, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12044, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---

 strace -z -f -o openat.log -e trace=open,openat python3 -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)'
-1
3000
grep openat openat.log | wc -l
402

This PR

time python -c "import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)"
-1
3000

real	0m0.805s
user	0m0.612s
sys	0m0.191s

nm -C $ROOTSYS/lib/*.so | grep TString::kNPOS | wc -l
1
nm -C $ROOTSYS/lib/*.so | grep kError | wc -l
27

strace -e file python -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)' 2>&1 | awk -v FS='"' '{ print $1 }' | sort | uniq -c
      1 -1
      1 3000
    219 access(
      2 chdir(
      1 execve(
      1 +++ exited with 0 +++
    184 getcwd(
      2 newfstatat(0, 
      2 newfstatat(1, 
      2 newfstatat(2, 
    532 newfstatat(3, 
     29 newfstatat(4, 
  10141 newfstatat(AT_FDCWD, 
    546 openat(AT_FDCWD, 
    843 readlink(
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12243, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12245, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---
      1 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=12249, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---

strace -z -f -o openat.log -e trace=open,openat python3 -c 'import ROOT;print(ROOT.gErrorIgnoreLevel);print(ROOT.kError)'
-1
3000
grep openat openat.log | wc -l
402

vepadulano · 2023-12-22T08:48:53Z

@vgvassilev It seems that the changes in this PR provide the same performance as the current master (after the two hotfixes), while supporting leaving the k* constants in TError.h as constexpr so without producing a symbol in libCore.so. For TString::kNPOS I guess we want to leave the commit there since it was anyway a real fix to a bug.

Another problem which is still present in both the current master and even after this PR is that the number of file operations is still quite high, i.e.

    532 newfstatat(3, 
     29 newfstatat(4, 
  10141 newfstatat(AT_FDCWD, 
    546 openat(AT_FDCWD, 
    843 readlink(

which are exactly the same whether current master or this patch. Consider that many of those newfstatat calls are spurious and I'm trying to understand where they come from. Take the following as an example

newfstatat(AT_FDCWD, "/home/vpadulan/programs/rootproject/rootbuild/reduce-symbol-search-release/gui/sessionviewer/input_line_11", 0x7ffde442f050, 0) = -1 ENOENT (No such file or directory)

Note the presence of the rootbuild directory, which is the path where I store the ROOT build artifacts, even though I am using ROOT from the install directory. I thought it was something in the machinery of (T)cling but so far my tests haven't brought me to any conclusion. Another suspicion is that the path to the build directory is hardcoded somehow in the .pcm or in the AST and doesn't get converted to the install directory.

vgvassilev · 2023-12-22T09:13:49Z

@vgvassilev It seems that the changes in this PR provide the same performance as the current master (after the two hotfixes), while supporting leaving the k* constants in TError.h as constexpr so without producing a symbol in libCore.so.

Thanks for doing this. That sounds good

For TString::kNPOS I guess we want to leave the commit there since it was anyway a real fix to a bug.

I agree, we should also land this PR since it supports now constexpr constructs which can be anywhere in external codes.

Consider that many of those newfstatat calls are spurious and I'm trying to understand where they come from.

These are likely garbage that's put by rootcling into the pcm files. We can verify if that theory is right if we make a build with -Druntime_cxxmodules=Off and see if we still get them at such high numbers...

smuzaffar · 2023-12-24T13:58:34Z

sure, started PR tests at cms-sw#193

dpiparo · 2023-12-27T06:48:37Z

Hi @smuzaffar , looking at the CMSSW tests, it seems this patch worked for CMS: could you confirm that?

dpiparo

Thanks @vgvassilev for this fix, delivered so quickly. I propose to merge the code if other reviewers approve the PR and if 2 conditions are met:

We get the confirmation from CMS that this is all what they need.
We understand why roottest-python-basic-basic fails on fedora38 and, if possible, fix the problem.

dpiparo · 2023-12-27T06:51:13Z

core/metacling/src/TCling.cxx

   DeclId_t d;
   TClingClassInfo *cl = (TClingClassInfo*)opaque_cl;

+   // Could trigger deserialization of decls.


I do not understand exactly why this had to move up in the cxx file: could you perhaps clarify, please?

Because the tests make GetDataMember now deserialize.

It seems I can just remove this and tests still pass on my system. Can you assess if this is still needed in the latest version of the PR? Shall we try reverting it?

No, it was needed on my system.

dpiparo · 2023-12-27T07:02:52Z

(btw, since we are looking for ways to speedup ROOT startup, I think this could also be an avenue #7123)

smuzaffar · 2023-12-27T08:30:14Z

Hi @smuzaffar , looking at the CMSSW tests, it seems this patch worked for CMS: could you confirm that?

Yes @dpiparo , this worked for CMS too. Local tests also show that with this update CMSSW/ROOT symbols mentioned in cms-sw/cmssw#43077 (comment) also do not trigger loading of all shared libs.

core/metacling/src/TClingCallbacks.cxx

core/metacling/src/TClingDataMemberInfo.cxx

interpreter/cling/lib/Interpreter/IncrementalExecutor.cpp

core/foundation/inc/TError.h

hahnjo · 2023-12-27T12:23:14Z

We understand why roottest-python-basic-basic fails on fedora38 and, if possible, fix the problem.

I believe this is not related to this PR, some of the Python tests fail from time to time on Fedora 38 and Fedora 39. I suspect there are still GC-related issues with Python 3.11 and 3.12 that were never fully understood...

phsft-bot · 2024-01-08T10:26:23Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

dpiparo · 2024-01-09T09:41:45Z

Hi @vgvassilev , thanks once again for this great development, which was delivered and tested on such a short timescale, basically during the end of the year break. On January 23rd, the last open pre-release of CMSSW_14_0_0 will close: after that date, no ROOT version updates will be possible (see cms-sw/cmssw#43077 (comment)).
Currently, the CMS issue has been worked around, therefore ROOT 6.30 can be picked up by CMS. However, it would be much better to have your fix in master and the 6.30 branch for CMS to pick up. That leaves us with a short window of opportunity, which is basically this week.

Do you think you can factor and succinctly document the changes in the way @hahnjo proposed in order to be able to merge the PR to make your fixes available to CMS? We see LHCb experiencing similar problems - they would also likely benefit from your code.

vgvassilev · 2024-01-09T11:01:15Z

Hi @vgvassilev , thanks once again for this great development, which was delivered and tested on such a short timescale, basically during the end of the year break. On January 23rd, the last open pre-release of CMSSW_14_0_0 will close: after that date, no ROOT version updates will be possible (see cms-sw/cmssw#43077 (comment)). Currently, the CMS issue has been worked around, therefore ROOT 6.30 can be picked up by CMS. However, it would be much better to have your fix in master and the 6.30 branch for CMS to pick up. That leaves us with a short window of opportunity, which is basically this week.

Do you think you can factor and succinctly document the changes in the way @hahnjo proposed in order to be able to merge the PR to make your fixes available to CMS? We see LHCb experiencing similar problems - they would also likely benefit from your code.

@dpiparo, I have made the changes that I agree with and responded to the rest of the comments. This PR seems to be ready to go from what concerns me.

The llvm9 JIT issued callbacks when a symbol was missing and we reacted on it by loading the relevant library. In root-project/root@9b2041e3 we have kept the logic but now the JIT started querying more often even for symbols which are okay to be missing. In turn that leads to scanning all libraries causing performance issues. This patch tries to limit this functionality only in contexts where automatic loading is allowed.

Co-authored-by: Jonas Hahnfeld <[email protected]>

Symbol lookup is a quite expensive operation and might result in JIT compilation and library loading. Co-authored-by: Jonas Hahnfeld <[email protected]>

This reverts commit 329fb5a.

dpiparo · 2024-01-11T10:21:11Z

Hi @hahnjo , @vgvassilev : thanks a lot for the exchanges and work on this topic. I was wondering if you can estimate an ETA for the merging of this nice work, key for addressing CMS (and LHCbs) current issues in a sustainable way and for the adoption of ROOT 6.30 by CMS in 2024.

phsft-bot · 2024-01-11T11:43:09Z

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

hahnjo

In any case, the organization is now good from my point of view.

hahnjo · 2024-01-11T16:03:04Z

core/metacling/src/TCling.cxx

   DeclId_t d;
   TClingClassInfo *cl = (TClingClassInfo*)opaque_cl;

+   // Could trigger deserialization of decls.


It seems I can just remove this and tests still pass on my system. Can you assess if this is still needed in the latest version of the PR? Shall we try reverting it?

dpiparo

Dear @vgvassilev , Thanks for these changes. Can you merge them asap? We can then make sure @smuzaffar uses this commit for the CMSSW builds in order for CMS to fully benefit from the improvements.

vgvassilev requested review from Axel-Naumann and pcanal as code owners December 20, 2023 17:39

pcanal requested review from dpiparo and hahnjo December 20, 2023 17:54

martamaja10 assigned vgvassilev Dec 21, 2023

martamaja10 added the in:Core Libraries label Dec 21, 2023

vgvassilev force-pushed the reduce-symbol-search branch from 99d82a3 to e95dd7f Compare December 21, 2023 21:20

vgvassilev force-pushed the reduce-symbol-search branch from e95dd7f to 250de1b Compare December 21, 2023 21:27

vgvassilev force-pushed the reduce-symbol-search branch from 250de1b to 9758c2e Compare December 22, 2023 07:24

vgvassilev force-pushed the reduce-symbol-search branch from 9758c2e to 5444a38 Compare December 22, 2023 07:53

dpiparo approved these changes Dec 27, 2023

View reviewed changes

hahnjo reviewed Dec 27, 2023

View reviewed changes

vgvassilev force-pushed the reduce-symbol-search branch from e2af2f9 to 8f4c26d Compare January 8, 2024 10:26

vgvassilev and others added 4 commits January 11, 2024 10:27

[core] Evaluate initializer of constant variables

c22be55

Co-authored-by: Jonas Hahnfeld <[email protected]>

[core] Prefer variable initializer over symbol search

566e73c

Symbol lookup is a quite expensive operation and might result in JIT compilation and library loading. Co-authored-by: Jonas Hahnfeld <[email protected]>

Revert "[core] Materialize symbols for TError variables"

55145f0

This reverts commit 329fb5a.

hahnjo force-pushed the reduce-symbol-search branch from 8f4c26d to 55145f0 Compare January 11, 2024 11:43

hahnjo approved these changes Jan 11, 2024

View reviewed changes

dpiparo self-requested a review January 15, 2024 15:44

dpiparo approved these changes Jan 15, 2024

View reviewed changes

vgvassilev merged commit 9ef2b9d into root-project:master Jan 15, 2024

This was referenced Jan 15, 2024

[core] Make some constants constexpr #14352

Merged

Cling triggers a huge number of openat calls when loading libraries #14277

Closed

v6-30 Backport cling fixes for CMS #14358

Merged

smuzaffar mentioned this pull request Jan 17, 2024

[ROOT6] Updated root to tip of branch master cms-sw/cmsdist#8948

Merged

vgvassilev mentioned this pull request Feb 4, 2024

PyROOT / cppyy: potentially ineffcicient "member offset" on enumerators #10724

Open

ferdymercury mentioned this pull request Feb 15, 2024

[core] ACLiC: speed up building / loading: #8017

Open

hahnjo mentioned this pull request Nov 18, 2024

[core] Fix early return in AutoloadLibraryGenerator #16967

Merged

ferdymercury mentioned this pull request Sep 19, 2025

TClass::Init() does two lookups where one is enough #7123

Open

[core] Reduce symbol search only to when autoloading is enabled. #14287

[core] Reduce symbol search only to when autoloading is enabled. #14287

Uh oh!

Conversation

vgvassilev commented Dec 20, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

phsft-bot commented Dec 20, 2023

Uh oh!

pcanal commented Dec 20, 2023

Uh oh!

github-actions bot commented Dec 20, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Test Results

Uh oh!

vepadulano commented Dec 21, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Baseline: master@fbd6abb

Import time

strace file stats

openat stats with the two symbols related to #14277

After this patch

Import time

strace file stats

openat stats with the two symbols related to #14277

Uh oh!

vgvassilev commented Dec 21, 2023

Uh oh!

phsft-bot commented Dec 21, 2023

Uh oh!

vgvassilev commented Dec 21, 2023

Uh oh!

phsft-bot commented Dec 21, 2023

Uh oh!

phsft-bot commented Dec 21, 2023

Failing tests:

Uh oh!

phsft-bot commented Dec 21, 2023

Failing tests:

Uh oh!

phsft-bot commented Dec 21, 2023

Failing tests:

Uh oh!

phsft-bot commented Dec 21, 2023

Failing tests:

Uh oh!

phsft-bot commented Dec 21, 2023

Failing tests:

Uh oh!

phsft-bot commented Dec 22, 2023

Uh oh!

phsft-bot commented Dec 22, 2023

Failing tests:

Uh oh!

phsft-bot commented Dec 22, 2023

Uh oh!

phsft-bot commented Dec 22, 2023

Failing tests:

Uh oh!

vepadulano commented Dec 22, 2023

Updates

Baseline: master before #14261 and #14276

Master after the two hotfixes

This PR

Uh oh!

vepadulano commented Dec 22, 2023

Uh oh!

vgvassilev commented Dec 22, 2023

Uh oh!

smuzaffar commented Dec 24, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dpiparo commented Dec 27, 2023

Uh oh!

dpiparo left a comment

Choose a reason for hiding this comment

Uh oh!

dpiparo Dec 27, 2023

Choose a reason for hiding this comment

Uh oh!

vgvassilev commented Dec 20, 2023 •

edited

Loading

github-actions bot commented Dec 20, 2023 •

edited

Loading

vepadulano commented Dec 21, 2023 •

edited

Loading

smuzaffar commented Dec 24, 2023 •

edited

Loading

smuzaffar commented Dec 27, 2023 •

edited

Loading