GDALRasterSource gives segmentation fault when reading rasters with NBITS=1

[jterry64]

Describe the bug

When trying to read rasters using GDALRasterSource that have nbits=1 and CCITTFAX4 compression GDAL options, I get the following error:

# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00007f5618facf8a, pid=13276, tid=0x00007f55e9371700
#
# JRE version: OpenJDK Runtime Environment (8.0_252-b09) (build 1.8.0_252-b09)
# Java VM: OpenJDK 64-Bit Server VM (25.252-b09 mixed mode linux-amd64 compressed oops)
# Problematic frame:
# C  [libc.so.6+0x9af8a]  __memcpy_sse2_unaligned_erms+0x1fa

If I change to the same raster but without any special GDAL options, it works fine.

To Reproduce

Here's a sample of a raster file that causes the issue: is__umd_tree_cover_gain_20N_100E.tif.zip

This how I'm using the GDALRasterSource and LayoutTileSource in my code:

https://github.com/wri/gfw_forest_loss_geotrellis/blob/3947899c4b985e9fad4f57bb7cf18bfc8fb279af/src/main/scala/org/globalforestwatch/layers/Layer.scala#L166

https://github.com/wri/gfw_forest_loss_geotrellis/blob/3947899c4b985e9fad4f57bb7cf18bfc8fb279af/src/main/scala/org/globalforestwatch/layers/Layer.scala#L314

I'm reading raster tiles from S3.

Expected behavior

No segmentation fault.

Environment

• Java version: 8
• Scala version: 2.12.6
• GeoTrellis version: 3.5.0
• GDAL version: 3.1.2

Additional context

Discussed with @pomadchin in gitter for context

[jterry64]

Hey @pomadchin, just checking in since it's been a few months - is there a plan to fix this at some point in the future?

[pomadchin]

Hi @jterry64 unfortunately we didn't have a chance to look into it. I can confirm that this bug is not present in Py bindings though, so that should be smth very specific to our bindings.

[pomadchin]

Hey @jterry64, could you verify that #3368 solves your issue?

Be careful with its testing though. The demo file size is only 8.4M, however, it is a 40000 x 40000 raster. Meaning that it would require a 40000 x 40000 Byte array allocation in JVM.

[pomadchin]

This issue was closed, but feel free to reopen if it is still an issue. Shortly the new 3.5.3-SNAPSHOT release would be available in our eclipse snapshots repo. That would simplify testing!