Improve saving geometry to root file

[lobis]

Currently the logic to save the TGeoManager into the root file is the following:

    ////////// Writing the geometry in TGeoManager format to the ROOT file
    ////////// Need to fork and do it in child process, to prevent occasional seg.fault
    pid_t pid;
    pid = fork();
    if (pid < 0) {
        perror("fork error:");
        exit(1);
    }
    // child process
    if (pid == 0) {
        // writing the geometry object
        freopen("/dev/null", "w", stdout);
        freopen("/dev/null", "w", stderr);

        REST_Display_CompatibilityMode = true;

        // We wait the father process ends properly
        sleep(5);

        // Then we just add the geometry
        auto file = new TFile(filename, "update");
        TGeoManager* geoManager = gdml->CreateGeoManager();

        file->cd();
        geoManager->SetName("Geometry");
        geoManager->Write();
        file->Close();
        exit(0);
    }
    // father process
    else {
        int stat_val = 0;
        pid_t child_pid;

        printf("Writing geometry ... \n");
    }

I find this overly complex for what it is trying to achieve, I think we should simplify it.

In principle something similar to the following code could be used:

    TGeoManager* geoManager = gdml->CreateGeoManager();
    if (geoManager) {
        geoManager->Write("Geometry");
        delete geoManager;
    } else {
        cout << "Error: could not create geometry" << endl;
    }

However if this change is implemented, ocassional segfaults will appear, as the description on the current implementation describes.

I think we should fix the underlying issue which is likely on the core repository.

Perhaps @nkx111 which I think authored the original implementation can describe how it works, and why a fork is necessary.

@juanangp or @jgalan perhaps you also have some insights on this.

[juanangp]

Can you try to remove delete geoManager; it might happen that you erase the pointer while it is still trying to write it, also memory management of root is not clear to me.

[lobis]

Can you try to remove delete geoManager; it might happen that you erase the pointer while it is still trying to write it, also memory management of root is not clear to me.

I tried to remove it and it didn't appear to have any effect. root memory management is not clear to me either but I think the problem is on our end, since I have used this same logic to save geometry in another project (https://github.com/lobis/radiation-transport/blob/main/applications/simulation/src/GlobalManager.cpp#L144) and it works without segfaults.

[nkx111]

Perhaps @nkx111 which I think authored the original implementation can describe how it works, and why a fork is necessary.

To some unknown reason the final step saving geometry to root file will crash occasionally. It happens randomly with about 3% probablity. So I added this fork step, to protect the main process, and make the pipeline always green.