Skip to content

Revert "[clang] survive -Inon-accessible-directory:" - #23758

Merged
hahnjo merged 1 commit into
root-project:masterfrom
hahnjo:revert-non-accessible
Nov 17, 2026
Merged

hahnjo merged 1 commit into
root-project:masterfrom
hahnjo:revert-non-accessible

Conversation

@hahnjo

@hahnjo hahnjo commented Oct 19, 2026 •

Copy link
Copy Markdown
Member

The Clang patch was added to run ROOT from non-accessible folders, see #12409. We cannot upstream the change because such include directories behave weirdly, and it is better to not support it.

This reverts commit b02e870.

@hahnjo hahnjo added this to the 6.42.00 milestone Oct 19, 2026
@hahnjo hahnjo self-assigned this Oct 19, 2026
@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown

Test Results

    21 files      21 suites   3d 12h 42m 31s ⏱️
 3 884 tests  3 884 ✅ 0 💤 0 ❌
72 201 runs  72 201 ✅ 0 💤 0 ❌

Results for commit daaa384.

♻️ This comment has been updated with latest results.

@hahnjo hahnjo removed this from the 6.42.00 milestone Nov 3, 2026
The Clang patch was added to run ROOT from non-accessible folders,
see root-project#12409. We cannot
upstream the change because such include directories behave weirdly,
and it is better to not support it.

This reverts commit b02e870.
@hahnjo
hahnjo force-pushed the revert-non-accessible branch from 34e4c93 to daaa384 Compare November 5, 2026 10:37
@hahnjo
hahnjo marked this pull request as ready for review November 5, 2026 10:39
@hahnjo

hahnjo commented Nov 5, 2026

Copy link
Copy Markdown
Member Author

FYI @ktf as previously communicated via email

@ktf

ktf commented Nov 5, 2026 via email

Copy link
Copy Markdown
Contributor

@hahnjo

hahnjo commented Nov 5, 2026

Copy link
Copy Markdown
Member Author

This breaks ALICE. I think you should provide an alternative solution, no?

This is why we communicated the intent to drop this patch in September. As explained, non-accessible folders behave very weirdly and also this patch is not acceptable upstream. In order to make progress, we believe the right course of action is to drop it.

In #12409, you already wrote

I am in parallel trying to understand if the unaccessible cwd can be avoided

I think this is what you should be doing on your side.

@ktf

ktf commented Nov 5, 2026

Copy link
Copy Markdown
Contributor

Indeed I though this was a different patch. I will check...

@hahnjo
hahnjo requested a review from dpiparo November 16, 2026 09:06
@hahnjo

hahnjo commented Nov 16, 2026

Copy link
Copy Markdown
Member Author

I would like to move forward with this revert. It is currently unable to distinguish between non-accessible directories and individual header files:

 $ cat TError.h 
#error ERROR
 $ ./bin/root.exe -q
In file included from input_line_4:2:
./TError.h:1:2: error: ERROR
#error ERROR
 ^
 $ chmod 0 TError.h
 $ ./bin/root.exe -q
   --------------------------------------------------------------------------------
  | Welcome to ROOT 6.43.01                                      https://root.cern |
  | (c) 1995-2026, The ROOT Team; conception: R. Brun, F. Rademakers               |
  | Built for linuxx8664gcc on Nov 16 2026, 09:12:07                               |
  | From heads/master@v6-41-99-123-g9ebc59dc5b9                                    |
  | With clang version 21.1.8 ( 21.1.8-1.module_el8.10.0+4172+b6b13d75) std201703  |
  | Try '.help'/'.?', '.demo', '.license', '.credits', '.quit'/'.q'                |
   --------------------------------------------------------------------------------

 $ echo $?
0

This is the situation that upstream was concerned about: just swallowing the error and continuing the search can be very confusing.

If still needed, I would like to better understand the ALICE deployment scenario and find a different solution (in ALICE or ROOT/Cling) than this "generic" downstream patch in Clang.

@ktf

ktf commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

IIRC, the deployment scenario was jobs without a $HOME during datataking. We need a solution for that.

@hahnjo

hahnjo commented Nov 16, 2026

Copy link
Copy Markdown
Member Author

IIRC, the deployment scenario was jobs without a $HOME during datataking. We need a solution for that.

Can you please share more information, so that it's possible to determine which solution is actually required? Even if running without $HOME, why does it need to be a completely non-accessible directory? Would it work to create a dedicated directory in /tmp?

But as I said, for me all these questions are orthogonal to the revert of this Clang patch, which I remain convinced should go away.

@ktf

ktf commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

running without a $HOME / $CWD was simply the minimal surface. Given we are redoing our Control system, I assume this can be accommodated.

@pcanal pcanal left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's go ahead

@hahnjo
hahnjo merged commit b380b7a into root-project:master Nov 17, 2026
58 of 60 checks passed
@hahnjo
hahnjo deleted the revert-non-accessible branch November 17, 2026 14:40
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants