Build ANTs on windows-2019 GitHub Actions CI runner
#3
Conversation
joshuacwnewton
force-pushed
the
jn/build-ants-on-windows
branch
from
16422c2 to
ba01bf9
Compare
| working-directory: antsbin | ||
| run: | | ||
| ls | ||
| make VERBOSE=1 -j 4 |
There was a problem hiding this comment.
During the make step, I ran into the exact same path-length issue that I ran into when building ITK locally on Windows (spinalcordtoolbox/spinalcordtoolbox#3682):
CMake Error at CMakeLists.txt:109 (message):
ITK source code directory path length is too long (64 > 50).Please move the
-- Configuring incomplete, errors occurred!
ITK source code directory to a directory with a shorter path.
See also "D:/a/spinalcordtoolbox-ants/spinalcordtoolbox-ants/antsbin/ITKv5-build/CMakeFiles/CMakeOutput.log".
As a short-term fix (as to not impede debugging), I'm going to rename the repository from spinalcordtoolbox-ants (22 chars) to build_ANTs (10 chars) which should save 12x2=24 characters, putting us well under the 50 char limit.
But, in the long-term, it would be nice to see if there are any better workarounds.
There was a problem hiding this comment.
There may be some hints on this github issue. In particular, there's a link to the ITK CMakeLists.txt file which includes the 50-character check, so it's an ITK-specific limit. It may be too conservative, since the Windows limit looks like it's 260 characters (does the ITK hierarchy require 260-50=210 characters?). It looks like the 50-character limit is not configurable, but there is an option ITK_SKIP_PATH_LENGTH_CHECKS which skips it entirely.
There was a problem hiding this comment.
!! Thank you for the sleuthing!
There was a problem hiding this comment.
Thank you from the future, @mguaypaq! Your comment here saved me some trouble while working on jqmcginnis/build_biomedia_stitching#1, as I needed to build ITK once more for that project. 😄
Specifying -DITK_SKIP_PATH_LENGTH_CHECKS=ON worked like a charm.
joshuacwnewton
force-pushed
the
jn/build-ants-on-windows
branch
from
5fc081d to
a4e7a97
Compare
joshuacwnewton
force-pushed
the
jn/build-ants-on-windows
branch
from
a4e7a97 to
ed5617a
Compare
This is an attempt to fix https://github.com/spinalcordtoolbox/build_ANTs/pull/3/files#r872663217, but may have unintended consequences for SCT. I'm just trying this out to see how the build goes across our platforms.
joshuacwnewton
force-pushed
the
jn/build-ants-on-windows
branch
from
89df670 to
07a466d
Compare
I don't think this is strictly necessary for our purposes.
…ces" This reverts commit 579502a.
joshuacwnewton
force-pushed
the
jn/build-ants-on-windows
branch
from
139a5c3 to
6849dc8
Compare
joshuacwnewton
changed the title
windows-2019 GitHub Actions CI runner
`build-ants.yml`: Skip ITK git clone in ANTs Superbuild
joshuacwnewton
force-pushed
the
jn/build-ants-on-windows
branch
from
a171bb6 to
3546114
Compare
|
I'm going to merge this PR, as the resulting binaries pass SCT's test suite. I plan on requesting further comments on the SCT pull request; I just want to keep the main discussion focused over there. |
| # snip out the apps we need for https://github.com/neuropoly/spinalcordtoolbox, since including all of ANTS is too much | ||
| mkdir sct-apps/ | ||
| cp antsbin/ANTS-build/Examples/Release/{antsRegistration,antsSliceRegularizedRegistration,antsApplyTransforms,ComposeMultiTransform}.exe sct-apps | ||
| (cd sct-apps; for i in `ls`; do mv $i isct_$i; done) |
There was a problem hiding this comment.
This is a nitpick, but for i in `ls` in bash is better written as for i in *, as per this article.
(This would have been out of scope for this PR anyway, since this pattern exists elsewhere in the file, and we shouldn't mess too much with something that's working. More of an FYI.)
This PR adds another workflow step to
build-ants.ymlto build ANTs for Windows.This would supersede the locally-built binaries from spinalcordtoolbox/spinalcordtoolbox#3682.