https://pantsbuild.org/ logo
Powered by
#general
  • e

    enough-analyst-54434

    09/05/2015, 12:02 AM
    so its a trap
  • w

    witty-crayon-22786

    09/05/2015, 12:03 AM
    unless it is functionally equivalent
  • w

    witty-crayon-22786

    09/05/2015, 12:03 AM
    which trap specifically are we talking about?
  • e

    enough-analyst-54434

    09/05/2015, 12:03 AM
    it rightly fails fast on the cases it can't handle
  • w

    witty-crayon-22786

    09/05/2015, 12:03 AM
    it is a trap for those who string concatenate rather than using os.path.join
  • w

    witty-crayon-22786

    09/05/2015, 12:03 AM
    is that supported behaviour with paths? i don't think so
  • e

    enough-analyst-54434

    09/05/2015, 12:03 AM
    then produces a different result from os.path.relpath for one of the cases it can
  • e

    enough-analyst-54434

    09/05/2015, 12:03 AM
    and it does not call that case out
  • e

    enough-analyst-54434

    09/05/2015, 12:04 AM
    I'll drop this , we disagree plain and simple
  • b

    bored-art-40741

    09/05/2015, 12:07 AM
    Copy code
    Traceback (most recent call last):
  File "/Users/pl/code/pants/src/python/pants/backend/jvm/tasks/jvm_compile/analysis_tools.py", line 66, in relativize
    self.parser.rebase_from_path(src_analysis, tmp_analysis_file, self._pants_home, self._PANTS_HOME_PLACEHOLDER, self._java_home)
  File "/Users/pl/code/pants/src/python/pants/backend/jvm/tasks/jvm_compile/analysis_parser.py", line 133, in rebase_from_path
    self.rebase(infile, outfile, pants_home_from, pants_home_to, java_home)
  File "/Users/pl/code/pants/src/python/pants/backend/jvm/tasks/jvm_compile/scala/zinc_analysis_parser.py", line 55, in rebase
    self._underlying_parser.rebase(infile, outfile, pants_home_from, pants_home_to, java_home)
  File "/Users/pl/code/pants/build-support/pants_dev_deps.venv/lib/python2.7/site-packages/zincutils/zinc_analysis_parser.py", line 88, in rebase
    rebase_element(Relations)
  File "/Users/pl/code/pants/build-support/pants_dev_deps.venv/lib/python2.7/site-packages/zincutils/zinc_analysis_parser.py", line 85, in rebase_element
    self._rebase_section(cls, header, infile, outfile, pants_home_from, pants_home_to, java_home)
  File "/Users/pl/code/pants/build-support/pants_dev_deps.venv/lib/python2.7/site-packages/zincutils/zinc_analysis_parser.py", line 117, in _rebase_section
    rebased_line = line.replace(pants_home_from, pants_home_to)
UnicodeDecodeError: 'ascii' codec can't decode byte 0xce in position 117: ordinal not in range(128)
  • r

    rough-minister-58256

    09/05/2015, 12:07 AM
    why not just name the specialized version something different to imply its not intended to be perfect reproduction of relpath?
  • e

    enough-analyst-54434

    09/05/2015, 12:07 AM
    it is named differently
  • e

    enough-analyst-54434

    09/05/2015, 12:07 AM
    fast_relpath
  • r

    rough-minister-58256

    09/05/2015, 12:07 AM
    i mean other than ‘relpath'
  • w

    witty-crayon-22786

    09/05/2015, 12:07 AM
    it executes the same operation, it just say in its doc that it doesn't abs or norm the inputs/outputs
  • e

    enough-analyst-54434

    09/05/2015, 12:08 AM
    I would just never have parsed its current pydic as meaning it gives different results than the stdlib relpath for stuff in the same tree
  • e

    enough-analyst-54434

    09/05/2015, 12:08 AM
    That could be a personal interpretation problem though
  • r

    rough-minister-58256

    09/05/2015, 12:08 AM
    to_relative_path etc
  • r

    rough-minister-58256

    09/05/2015, 12:08 AM
    would make it clear
  • e

    enough-analyst-54434

    09/05/2015, 12:09 AM
    it would also be performant and clear to just make it work the same unless I'm missing something
  • r

    rough-minister-58256

    09/05/2015, 12:09 AM
    or that too simple smile
  • e

    enough-analyst-54434

    09/05/2015, 12:10 AM
    none of us saw the difference but it seems a bit odd to wantonly keep the difference if it can be removed now that we know about it
  • r

    rough-minister-58256

    09/05/2015, 12:10 AM
    true
  • b

    bored-art-40741

    09/05/2015, 12:12 AM
    I think I have a fix for the name hashing issue
  • b

    bored-art-40741

    09/05/2015, 12:13 AM
    yup I do
  • b

    bored-art-40741

    09/05/2015, 12:13 AM
    And it's on the pants side of things
  • w

    witty-crayon-22786

    09/05/2015, 12:14 AM
    woot
  • b

    bored-art-40741

    09/05/2015, 12:14 AM
    You're going to love this
  • b

    bored-art-40741

    09/05/2015, 12:14 AM
    Copy code
    $ git diff
diff --git a/src/python/pants/backend/jvm/tasks/jvm_compile/analysis_tools.py b/src/python/pants/backend/jvm/tasks/jvm_compile/analysis_tools.py
index a0fcb1c..0da7663 100644
--- a/src/python/pants/backend/jvm/tasks/jvm_compile/analysis_tools.py
+++ b/src/python/pants/backend/jvm/tasks/jvm_compile/analysis_tools.py
@@ -19,12 +19,12 @@ class AnalysisTools(object):
   # for that reason.
   # TODO: If some future change requires us to invalidate all cached artifacts for some good reason
   # (by bumping GLOBAL_CACHE_KEY_GEN_VERSION), we can use that opportunity to change this string.
-  _PANTS_HOME_PLACEHOLDER = '/_PANTS_HOME_PLACEHOLDER'
+  _PANTS_HOME_PLACEHOLDER = b'/_PANTS_HOME_PLACEHOLDER'

   def __init__(self, java_home, parser, analysis_cls):
     self.parser = parser
     self._java_home = java_home
-    self._pants_home = get_buildroot()
+    self._pants_home = get_buildroot().encode('utf-8')
     self._analysis_cls = analysis_cls

   def split_to_paths(self, analysis_path, split_path_pairs, catchall_path=None):
  • w

    witty-crayon-22786

    09/05/2015, 12:14 AM
    ... hm
1...299300301...2420Latest