How do we check if a file is corrupted?

[acampove]

Feature description

Hi,

I have a collection of 500 files and I need to run a sanity check. However I cannot find anything like:

import ROOT

status = ROOT.TFile.check_health('file.root')

if status != 0:
    raise ValueError('message here')

How is the user supposed to know if a file has problems among hundreds?

Alternatives considered

No response

Additional context

No response

[wacfrr]

Regarding the integrity check for this batch of ROOT files, I suggest we handle it with the following steps:

- Read the file list: First, automatically read the list of all files in the target directory.

- Iterate through the files: Loop through each file and perform the following operations:
    - Try to open and check: Attempt to open the file within the loop and determine whether it is a zombie (corrupted file).

    - Catch and print exceptions: Wrap this process in an exception-handling block to catch all exceptions if something goes wrong, and print the specific error messages so we can precisely pin down which files have issues.

    - Ignore and proceed: Once an error is caught, simply ignore the exception and continue running until all files are checked, ensuring the script won't crash halfway through.

[ferdymercury]

if a file has problems

TFile::IsZombie

Or does that not detect some corruption cases?

[acampove]

if a file has problems

TFile::IsZombie

Or does that not detect some corruption cases?

I am not sure what's going on here, are you asserting, as a ROOT maintainer, that IsZombie will be enough to guarantee that the file is not corrupted? Or are you asking me, a ROOT user, that question? As a ROOT user, I do not know, that's why I opened this issue.

[dpiparo]

Thanks for this report. I suggest we keep a positive register at all times in our messages. This is written text, and sometimes the tone may not be obvious :-)

[ferdymercury]

that IsZombie will be enough to guarantee that the file is not corrupted?

That is what the online documentation of that function claims.

Or are you asking me, a ROOT user, that question?

My suggestion was that you try that function locally, if it does not work as advertised, please report it here / reopen this issue.

[vepadulano]

A simple way to check if something went wrong when opening the file is as follows

#include <TFile.h>
#include <memory>

int main()
{
    auto f = std::unique_ptr<TFile>{TFile::Open("myfile.root")};
    if (
        f->IsZombie() || // Something went wrong during TFile construction 
        f->TestBit(TFile::kRecovered) // The TFile had to run the recovery mechanism when opening the file
        )
        throw std::runtime_error("Something went wrong when opening the ROOT file.");
}

Also https://root.cern.ch/doc/v640/classTFile.html#ad0377adf2f3d88da1a1f77256a140d60 and https://root.cern.ch/doc/v640/classTFile.html#ae5f85941cbcb48048de3b8d2df24540b

[ferdymercury]

A PR improving the documentation is here: #22477