Re: [lmi] Question about using fardels

[Greg Chicares]

On 2020-09-21 21:53, Vadim Zeitlin wrote:
> 
>  After using "make fardel" for performance testing, as shown in your
> emails, I've realized that I don't actually understand how is it supposed
> to work: it seems to create all the files in a single directory, unlike a
> normal lmi installation which uses separate "bin" and "data" directories,
> and using this directory itself as "data" doesn't work for me, as it
> results in an error message:
> 
>       % ./lmi_cli_shared.exe --accept --data_path=. --selftest
>       Value 'sample2naic' invalid for type 'ce_product_name'.
>       [ce_product_name.cpp : 145]

I've never seen anything like that, so I'm sure I won't be able
to reproduce it. Let's see...

/tmp/lmi-20200921T1150Z[126]$wine ./lmi_cli_shared.exe --accept --data_path=. 
--selftest
Value 'sample2naic' invalid for type 'ce_product_name'.
[ce_product_name.cpp : 145]

Oh. Hmmm.

Wait...the message really means what it says, once you see
what it means: there are no 'sample2naic' product files on
the data path. To fix that, brutally [but see below instead]:

/tmp/lmi-20200921T1150Z[126]$wine ./product_files.exe
Generating product files.

All product files written.

/tmp/lmi-20200921T1150Z[0]$wine ./lmi_cli_shared.exe --accept --data_path=. 
--selftest
Test speed:
[...continues normally]

>  I can make it work by using data directory created by installing the
> latest master version,

Yes, the command I'd been using was:

/tmp/lmi-20200921T1150Z[0]$wine /opt/lmi/bin/lmi_cli_shared.exe --accept 
--data_path=/opt/lmi/data --selftest

> but I'd still like to understand: aren't fardels not
> supposed to be self-contained? So perhaps I'm just using them wrong? But
> what is the right way to use them then?

You're using them the right way. And if you pull commit b99f6ac4f5498,
the right way will actually work. You can use '--data_path=.' if you
want, or you can just omit that and let it default to '.'.

Thanks for reporting this defect.