fix glob dir on windows #8654
fix glob dir on windows #8654
Conversation
danini-the-panini
force-pushed
the
ds-fix-dir-glob-windows
branch
from
24c51ad to
6f3b3b1
Compare
danini-the-panini
force-pushed
the
ds-fix-dir-glob-windows
branch
from
6f3b3b1 to
0a0b745
Compare
enebo
left a comment
enebo
left a comment
There was a problem hiding this comment.
I believe we should be using the String construction listed above in my comment for your createResource line. I think links.size() > 0 section also uses the same call and it should also be changed.
I am going to ask @headius on channel to review this since glob is always too complicated.
| buf.append( isRoot(base) ? EMPTY : SLASH ); | ||
| buf.append( getBytesInUTF8(file) ); | ||
| if ( SLASH_INDEX == -1 ) { | ||
| resource = JRubyFile.createResource(runtime, cwd, RubyString.byteListToString(buf)); |
There was a problem hiding this comment.
Up above we see the same resource creation but it uses the encoding for charset:
resource = JRubyFile.createResource(runtime, cwd, new String(buf.unsafeBytes(), buf.begin(), buf.length(), enc.getCharset()));This enc comes from the original pattern and that pattern is fed through RubyFile#get_path/filePathConvert which has special logic for transcoding if NOT on windows to match filesystem encoding. I think the use of byteListToString (used both here and below) is wrong. It will make an ISO-8859-1 string where each byte is a char vs the line above which properly decodes the bytes given with the charset provided.
I think the reason we do not ever see problems is almost all files are single-byte ASCII.
There was a problem hiding this comment.
Yes I think this risks losing encoding and possibly dealing with mangled characters. If we cannot fix it by working with the possibly-multibyte characters directly, then we should use the functions in RubyEncoding to decode them to String properly.
| resource = JRubyFile.createResource(runtime, cwd, RubyString.byteListToString(link)); | ||
| if ( resource.isDirectory() ) { | ||
| final int len = link.getRealSize(); | ||
| buf.length(0); | ||
| buf.append(link); | ||
| buf.append(path, SLASH_INDEX, end - SLASH_INDEX); | ||
| status = glob_helper(runtime, cwd, scheme, buf, buf.getBegin() + len, flags, func, arg); | ||
| } | ||
| final int len = link.getRealSize(); | ||
| buf.length(0); | ||
| buf.append(link); | ||
| buf.append(path, SLASH_INDEX, end - SLASH_INDEX); | ||
| status = glob_helper(runtime, cwd, scheme, buf, buf.getBegin() + len, flags, func, arg); |
There was a problem hiding this comment.
so i moved the isDirectory check to where we append to links rather than the loop. This might cause an issue where links would have contained only non-directories, it should still call break mainLoop, although I haven't seen any issues. I don't know how I would cause this to happen.
There was a problem hiding this comment.
I've reverted this change
danini-the-panini
force-pushed
the
ds-fix-dir-glob-windows
branch
from
86cbebf to
03945c6
Compare
3 participants
Resolve #8623