LHAPDF path incorrectly set in 2.6.0

[Zachary Marshall]

Hi,

In MadGraph5_aMC@NLO 2.6.0, in madgraph/interface/common_run_interface.py on L3963 the get_lhapdf_pdfsetsdir function is defined. This function includes:

        # check if the LHAPDF_DATA_PATH variable is defined
        if 'LHAPDF_DATA_PATH' in os.environ.keys() and os.environ['LHAPDF_DATA_PATH']:
            datadir = os.environ['LHAPDF_DATA_PATH']

This path is allowed to be a colon-separated series of paths (just like any other $PATH in the environment). That causes crashes when running with certain environments. Ultimately this leads to a crash on L3930 of madgraph/interface/common_run_interface.py:

            pdfsets_lines = \
                    [l for l in open(pjoin(pdfsets_dir, 'pdfsets.index')).read().split('\n') if l.strip()]

That needs to be able to handle the colon-separated path, and just search through for the first pdfsets.index file it finds.

Thanks,
Zach

[Olivier Mattelaer (olivier-mattelaer)]

Thanks,

I will push the following patch:

=== modified file 'madgraph/interface/common_run_interface.py'
--- madgraph/interface/common_run_interface.py 2017-10-23 22:48:09 +0000
+++ madgraph/interface/common_run_interface.py 2017-11-09 16:26:15 +0000
@@ -4041,7 +4041,14 @@
         elif lhapdf_version.startswith('6.'):
             datadir = subprocess.Popen([self.options['lhapdf'], '--datadir'],
                          stdout = subprocess.PIPE).stdout.read().strip()
-
+
+ if ':' in datadir:
+ for totry in datadir.split(':'):
+ if os.path.exists(pjoin(totry, 'pdfsets.index')):
+ return totry
+ else:
+ return None
+
         return datadir

     def get_lhapdf_libdir(self):

Cheers,

Olivier

> On Nov 9, 2017, at 09:20, Zachary Marshall <email address hidden> wrote:
>
> Public bug reported:
>
> Hi,
>
> In MadGraph5_aMC@NLO 2.6.0, in
> madgraph/interface/common_run_interface.py on L3963 the
> get_lhapdf_pdfsetsdir function is defined. This function includes:
>
> # check if the LHAPDF_DATA_PATH variable is defined
> if 'LHAPDF_DATA_PATH' in os.environ.keys() and os.environ['LHAPDF_DATA_PATH']:
> datadir = os.environ['LHAPDF_DATA_PATH']
>
> This path is allowed to be a colon-separated series of paths (just like
> any other $PATH in the environment). That causes crashes when running
> with certain environments. Ultimately this leads to a crash on L3930 of
> madgraph/interface/common_run_interface.py:
>
> pdfsets_lines = \
> [l for l in open(pjoin(pdfsets_dir, 'pdfsets.index')).read().split('\n') if l.strip()]
>
> That needs to be able to handle the colon-separated path, and just
> search through for the first pdfsets.index file it finds.
>
> Thanks,
> Zach
>
> ** Affects: mg5amcnlo
> Importance: Undecided
> Status: New
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1731176
>
> Title:
> LHAPDF path incorrectly set in 2.6.0
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> In MadGraph5_aMC@NLO 2.6.0, in
> madgraph/interface/common_run_interface.py on L3963 the
> get_lhapdf_pdfsetsdir function is defined. This function includes:
>
> # check if the LHAPDF_DATA_PATH variable is defined
> if 'LHAPDF_DATA_PATH' in os.environ.keys() and os.environ['LHAPDF_DATA_PATH']:
> datadir = os.environ['LHAPDF_DATA_PATH']
>
> This path is allowed to be a colon-separated series of paths (just
> like any other $PATH in the environment). That causes crashes when
> running with certain environments. Ultimately this leads to a crash
> on L3930 of madgraph/interface/common_run_interface.py:
>
> pdfsets_lines = \
> [l for l in open(pjoin(pdfsets_dir, 'pdfsets.index')).read().split('\n') if l.strip()]
>
> That needs to be able to handle the colon-separated path, and just
> search through for the first pdfsets.index file it finds.
>
> Thanks,
> Zach
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.n...

[Zachary Marshall (zach-marshall)]

Thanks Olivier! Can I ask what the timescale for the next release is? I just want to understand whether we should apply this privately or wait for 2.6.1.

Best,
Zach

[Olivier Mattelaer (olivier-mattelaer)]

Probably one or two week.

Olivier
> On Nov 9, 2017, at 19:05, Zachary Marshall <email address hidden> wrote:
>
> Thanks Olivier! Can I ask what the timescale for the next release is?
> I just want to understand whether we should apply this privately or wait
> for 2.6.1.
>
> Best,
> Zach
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1731176
>
> Title:
> LHAPDF path incorrectly set in 2.6.0
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> In MadGraph5_aMC@NLO 2.6.0, in
> madgraph/interface/common_run_interface.py on L3963 the
> get_lhapdf_pdfsetsdir function is defined. This function includes:
>
> # check if the LHAPDF_DATA_PATH variable is defined
> if 'LHAPDF_DATA_PATH' in os.environ.keys() and os.environ['LHAPDF_DATA_PATH']:
> datadir = os.environ['LHAPDF_DATA_PATH']
>
> This path is allowed to be a colon-separated series of paths (just
> like any other $PATH in the environment). That causes crashes when
> running with certain environments. Ultimately this leads to a crash
> on L3930 of madgraph/interface/common_run_interface.py:
>
> pdfsets_lines = \
> [l for l in open(pjoin(pdfsets_dir, 'pdfsets.index')).read().split('\n') if l.strip()]
>
> That needs to be able to handle the colon-separated path, and just
> search through for the first pdfsets.index file it finds.
>
> Thanks,
> Zach
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1731176/+subscriptions

[Zachary Marshall (zach-marshall)]

Hi Olivier,

Do you have a new estimate for the release time?

Thanks,
Zach

[Olivier Mattelaer (olivier-mattelaer)]

Hi Zach,

All the feature that was targeted for that version have now been validated and they all have been merged in a single version.
So the release will happen very soon. Maybe during the week-end (but a bit unlikely),
If we miss that time-slot, it will be for the end of next week —probably on thursday—
 (The reason for such gap is that I have only a 60% contract with the physics department and therefore I do work for some other department on Monday/Tuesday)

Cheers,

Olivier

> On Nov 30, 2017, at 13:27, Zachary Marshall <email address hidden> wrote:
>
> Hi Olivier,
>
> Do you have a new estimate for the release time?
>
> Thanks,
> Zach
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1731176
>
> Title:
> LHAPDF path incorrectly set in 2.6.0
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> In MadGraph5_aMC@NLO 2.6.0, in
> madgraph/interface/common_run_interface.py on L3963 the
> get_lhapdf_pdfsetsdir function is defined. This function includes:
>
> # check if the LHAPDF_DATA_PATH variable is defined
> if 'LHAPDF_DATA_PATH' in os.environ.keys() and os.environ['LHAPDF_DATA_PATH']:
> datadir = os.environ['LHAPDF_DATA_PATH']
>
> This path is allowed to be a colon-separated series of paths (just
> like any other $PATH in the environment). That causes crashes when
> running with certain environments. Ultimately this leads to a crash
> on L3930 of madgraph/interface/common_run_interface.py:
>
> pdfsets_lines = \
> [l for l in open(pjoin(pdfsets_dir, 'pdfsets.index')).read().split('\n') if l.strip()]
>
> That needs to be able to handle the colon-separated path, and just
> search through for the first pdfsets.index file it finds.
>
> Thanks,
> Zach
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1731176/+subscriptions

[Zachary Marshall (zach-marshall)]

Hi Olivier,

Sorry to be a pain, but I just want to double-check that 2.6.1 is on track for today -- we have many excited users champing at the bit for this one!

Thanks,
Zach