[io] Increase a file size tolerance for arm builds. #1196
Conversation
root/io/filemerger/CMakeLists.txt
Outdated
| @@ -340,7 +340,7 @@ if(${compression_default} STREQUAL "zlib") | |||
| if(CMAKE_SYSTEM_PROCESSOR MATCHES aarch64) | |||
| ROOTTEST_ADD_TEST(simple-lz4-compr-level9 | |||
| PRECMD ${ROOT_hadd_CMD} -f409 hsimple409.root hsimple.root | |||
| COMMAND ${ROOT_root_CMD} -q -l -b "${CMAKE_CURRENT_SOURCE_DIR}/testSimpleFile.C(\"hsimple409.root\",25000,409,516975,5)" | |||
| COMMAND ${ROOT_root_CMD} -q -l -b "${CMAKE_CURRENT_SOURCE_DIR}/testSimpleFile.C(\"hsimple409.root\",25000,409,516975,10)" | |||
There was a problem hiding this comment.
This is already in the aarch64 sections and thus the difference is from prior/different runs on arm64. Do we understand what changed?
There was a problem hiding this comment.
I have no idea what changed. It's the first run of the new alma9-arm runners.
git blame says:
Author: Oksana Shadura <[email protected]>
Date: Mon Jul 8 11:47:27 2019 +0200
Adjust file size for LZ4 (409 for aarch64)
Maybe the change is 5 years of evolution of distro and lz4. I'm not sure if this test ran in the mean time.
There was a problem hiding this comment.
Do you think it's safe to update the size? 7 out of 517k bytes seems like a small change.
There was a problem hiding this comment.
I updated to the commit message to be a bit clearer.
There was a problem hiding this comment.
Thanks. We should also clarify if it is used the system lz4 or the builtin version.
hageboeck
force-pushed
the
armCI
branch
2 times, most recently
from
1351bdf
to
393f0af
Compare
|
Can we also note the raw number (in particular to remember whether it is a slight increase or slight decrease in size :) )? Thanks. |
Sure, in the commit or in the CMakeLists? Can I ask, though, if changes on the level of 0.0052% are relevant? Shouldn't we be more forgiving, and only start looking when we reach something like the per-mille level? |
Commit log ;)
They can be. Changes that are 'problems' that can lead to small changes includes unintentional but consistent changes to the I/O (i.e. content has changed/increased but everything works thanks to schema evolution) .... And technically the unintentional change might be a bug/random-behavior that is being seen only rare platforms - so if the change was not so small and only in one file, the right process would have been to:
i.e. this PR assumes the change is only due to a change in the compression algorithm and is not related to a architecture specific flaw in the ROOT code. |
With alma9 and its system lz4 1.9.3, the filesize of the 409 test increased. 516975 --> 516703 byes. The previous value is from 2019 (unknown distro). Increased the tolerance to 30 now.
|
Thanks. |
On arm, a filesize comes out 27 bytes different from
x86the 2019 arm values, so the tolerance had to be increased a bit.This is a prerequisite for #root-project/root-16526