Hello, I'm reaching out to ask if there is a known error or if anything has changed within ENVI 5.5.3 + IDL 8.7.3 (64-bit) that would affect linear spectral unmixing (LSU) results between June 10, 2020 and July 15, 2020.

Since June 10, I have been working with the same endmembers (12 single pixels saved as 12 separate ROIs for bare soil and vegetation types), the same two spectrally and spatially subset Landsat 8 images, and I always apply a unit sum constraint of 1.0.  For each run, I output the file to Memory and then copy and paste the fractional cover results from the Cursor Value pop-up window to my personal Excel spreadsheet.

A week or so ago, I began receiving the follow error when I chose the "Memory" option for the LSU output file (and not every time, only occasionally):
"unmix_doit: An error has occurred during processing
Error: "Invalid pointer: )>."
The results may be invalid."

In order to get LSU working again, I completely exited out of ENVI and re-started the program.  I started saving the files to my hard drive, rather than to Memory.  Sometimes the "X Start" and "Y Start" cursor values would shift with the output image file, but I adjusted those in the metadata and did not worry (and changing X and Y Starts never appeared to impact LSU fractional cover output values).

Today, I masked a few additional pixels in the image to "remove" some water and impervious pixels.  Suddenly, I received new LSU output values.  The weirdest thing?  When I return the same images as from June 10, I now receive the same output values as those from the updated masked image file today, and I receive the same values with other spatial subsets of the same image data (meaning, I cannot recreate the output values from the attempts on June 10, and I only receive these new values, no matter the spatial subset).

To summarize, there are a few things that stand out to me that might (hopefully) help point you to a solution: (1) The occasional error message when saving to Memory (above), (2) the seemingly random change in X and Y Start values, (3) the newly masked pixels (although that should not be it), or (4) something on the ENVI side that has been buggy.

Thank you very much for your help!