hi guys,
I encountered an issue with File#size returning 20047 on windows by
spreadsheet gem, and I can reproduce the same error in jirb as below:
(latest jruby 1.7.3, windows 7, jdk1.6.0_26)
C:\Users\me>jirb
irb(main):001:0> f=File.new “foo.txt”
=> #<File:foo.txt>
irb(main):002:0> f.size
SystemCallError: Unknown error - Unknown Error (20047) - -1
from org/jruby/RubyFile.java:1082:in size' from (irb):2:in evaluate’
from org/jruby/RubyKernel.java:1066:in eval' from org/jruby/RubyKernel.java:1409:in loop’
from org/jruby/RubyKernel.java:1174:in catch' from org/jruby/RubyKernel.java:1174:in catch’
from C:\jruby\jruby-1.7.3\bin\jirb:13:in `(root)’
I found two bugs in jira about this issue:
The first one is :
[windows] File.size is completely broken on Windows
with the link http://jira.codehaus.org/browse/JRUBY-4820
the Status is “Resolved”, the Resolution is “Fixed”, but Fix Version/s
is “None”.
The second one is:
Changing jnr-posix to use errno() value unconditionally has created lots
of failures in windows
with the link http://jira.codehaus.org/browse/JRUBY-6136
the status is “Open”, Resolution is “Unresolved”.
So, do we still have this bug in latest v1.7.3?
I want to have a confirmation, then we can decide what to do next?
Or any workaround for this issue?
Thanks.
I re-opened the JRUBY-4820 to make sure we don’t lose it and based on
comments it makes me wonder if it has ever fully been fixed or not.
It is a really basic method but File.stat(file) probably uses ‘stat’
and File.size might use ‘fstat’ which could be some weird posix compat
issue on Windows for us (since I have not looked at these methods in
quite a while I will need to revisit this). Worst case we can
special case windows and use the jnr-posix method we know works for
Windows. Best case we solve what is wrong in jnr-posix to properly
support this.
The 20047 error is an interesting error because it still would have
been a bug for you. I am assuming foo.txt exists when you ran that.
So it should have worked…but didn’t…then returned a nonsense error
instead of the error of a ENOENT. So if you can somehow squeeze in a
third bug into this single call it will be a trifecta 
Color us embarrassed…
-Tom
PS- JRUBY-6136 is a difficult issue to solve in my mind. jnr-posix is
supposed to expose posix methods and we want proper granularity of
errno values. If we mask all the various windows errnos we will gain
more cross-platform consistency at the cost that consumers of
jnr-posix (jruby is not the only user of it) will lose the specific
errnos (who am I to say knowing a mapped network drive errno is not
what someone wants). At this point we should capture the two weird
errnos we know about on the JRuby side as an interim solution and
remap them to ENOENT.
On Mon, Mar 25, 2013 at 6:52 AM, h h [email protected] wrote:
from (irb):2:in `evaluate'
with the link http://jira.codehaus.org/browse/JRUBY-4820
Or any workaround for this issue?
Thanks.
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
–
blog: http://blog.enebo.com twitter: tom_enebo
mail: [email protected]
hi Thomas,
Thanks for your quick reply.
Your assumption is right, foo.txt exist here when I reproduced the
issue.
Thanks.
— 13年3月25日，周一, Thomas E Enebo [email protected] 写道：