#1293 -buffersize seems to do nothing (?)

obsolete: 8.2
closed
4
2001-09-13
2000-10-26
Anonymous
No

OriginalBugID: 2653 Bug
Version: 8.2
SubmitDate: '1999-08-31'
LastModified: '1999-11-07'
Severity: SER
Status: UnAssn
Submitter: techsupp
ChangedBy: hobbs
OS: Windows 95
Machine: X86
FixedDate: '2000-10-25'
ClosedDate: '2000-10-25'

Name:
keith lea

ObservedBehavior:
% fconfigure file56e9fc
-blocking 1 -buffering full -buffersize 1048576 -encoding binary
-eofchar {} -translation lf

and yet, when copying to this file from a socket, the disk buzzes
nonstop and the file's size increases steadily, instead of every
megabyte. I do not use [flush].

DesiredBehavior:
configuring a channel as "-buffersize 1048576" should not write to the
disk until the buffer reaches 1048576 bytes.

Discussion

  • Donal K. Fellows

    This is a known misfeature of the Windows file channel code; it flushes on every write. Apparently, this is necessary to get certain tests to behave in an expected way, but I reckon that it is probably better to rethink the tests instead... :^)

     
  • Donal K. Fellows

    • labels: 104247 --> 27. Channel Types
     
  • Andreas Kupries

    Andreas Kupries - 2001-07-19

    Logged In: YES
    user_id=75003

    This is not necessary a problem of the windows channel
    code. He claims that the write occurs before the buffer is
    full. This claim we will have to check.

     
  • Andreas Kupries

    Andreas Kupries - 2001-08-23

    Logged In: YES
    user_id=75003

    See Tcl Bug #219148 for the item Donal spoke of.

     
  • Andreas Kupries

    Andreas Kupries - 2001-08-23
    • assigned_to: nobody --> andreas_kupries
     
  • Andreas Kupries

    Andreas Kupries - 2001-08-24

    Logged In: YES
    user_id=75003

    Was something written to the file before the greater
    buffersize was configured ?

    It could be a case or reusing the small buffer from
    such a previous write.

     
  • Andreas Kupries

    Andreas Kupries - 2001-08-24

    Logged In: YES
    user_id=75003

    A script exercising the problem would be nice.

     
  • Andreas Kupries

    Andreas Kupries - 2001-09-07

    Logged In: YES
    user_id=75003

    A patch for Performance Bug #219148 (I/O for NT is 80-100
    times slower than for Unix) was submitted today to the 8.4
    and 8.3.3 branches. The submitter of this item is asked to
    check if that patch helps him in his situation too.

     
  • Andreas Kupries

    Andreas Kupries - 2001-09-07
    • priority: 5 --> 4
     
  • Andreas Kupries

    Andreas Kupries - 2001-09-11

    Logged In: YES
    user_id=75003

    keith lea = <keith@cs.oswego.edu>

     
  • Andreas Kupries

    Andreas Kupries - 2001-09-13
    • status: open --> closed
     

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks