# newinstall.sh should warn about env vars known to cause problems

XMLWordPrintable

## Details

• Type: Story
• Status: Done
• Resolution: Done
• Fix Version/s: None
• Component/s:
• Labels:
None
• Story Points:
1
• Team:
SQuaRE

## Activity

Hide
Joshua Hoblitt added a comment -

Jonathan Sick Is this approximately what you had in mind?

 WARNING: the following environment variables are defined that will effect the operation of the LSST build tooling.   LSST_HOME="foo" EUPS_PATH="bar" EUPS_PKGROOT="baz" REPOSITORY_PATH="qux"   It is recommend that they are undefined before running this script.   unset LSST_HOME EUPS_PATH EUPS_PKGROOT REPOSITORY_PATH 

Show
Joshua Hoblitt added a comment - Jonathan Sick Is this approximately what you had in mind? WARNING: the following environment variables are defined that will effect the operation of the LSST build tooling.   LSST_HOME= "foo" EUPS_PATH= "bar" EUPS_PKGROOT= "baz" REPOSITORY_PATH= "qux"   It is recommend that they are undefined before running this script.   unset LSST_HOME EUPS_PATH EUPS_PKGROOT REPOSITORY_PATH
Hide
Jonathan Sick added a comment -

Sure. Will this exit the script, prompt for confirmation to continue, or just continue the newinstall script anyways?

Show
Jonathan Sick added a comment - Sure. Will this exit the script, prompt for confirmation to continue, or just continue the newinstall script anyways?
Hide
Joshua Hoblitt added a comment -

The current implementation prints the warning early on and then continues. I thought that was the requested behavior on slack this morning?

Show
Joshua Hoblitt added a comment - The current implementation prints the warning early on and then continues. I thought that was the requested behavior on slack this morning?
Hide
Jonathan Sick added a comment -

I wonder if only a warning is insufficient? People will miss the warning in the stream and I'll still have to document it.

What do you think about having the warning block the script for confirmation (unless -b is used)?

Show
Jonathan Sick added a comment - I wonder if only a warning is insufficient? People will miss the warning in the stream and I'll still have to document it. What do you think about having the warning block the script for confirmation (unless -b is used)?
Hide
Joshua Hoblitt added a comment -

I'm fairly ambivalent about it. I'll make make the warning fatal if not in batch mode.

Show
Joshua Hoblitt added a comment - I'm fairly ambivalent about it. I'll make make the warning fatal if not in batch mode.
Hide
Jonathan Sick added a comment -

Perfect :+1:

Show
Jonathan Sick added a comment - Perfect :+1:
Hide
Joshua Hoblitt added a comment -

I've been looking through the lsst github org and I don't think there are any hard dependencies on newinstall.sh's loadLSST. files declaring LSST_HOME but I'm not 100% sure. However, I can't think of a reason why the loadLSST. files would need to respect a pre-declared LSST_HOME. I think it we remove that behavior, we don't need to doc or warn about setting LSST_HOME.

Show
Joshua Hoblitt added a comment - I've been looking through the lsst github org and I don't think there are any hard dependencies on newinstall.sh's loadLSST. files declaring LSST_HOME but I'm not 100% sure. However, I can't think of a reason why the loadLSST. files would need to respect a pre-declared LSST_HOME . I think it we remove that behavior, we don't need to doc or warn about setting LSST_HOME .
Hide
Joshua Hoblitt added a comment -

merged.

Show
Joshua Hoblitt added a comment - merged.

## People

• Assignee:
Joshua Hoblitt
Reporter:
Joshua Hoblitt
Watchers:
Jonathan Sick, Joshua Hoblitt