This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Increase width of padded output in stdio-common/tst-put-error
- From: David Miller <davem at davemloft dot net>
- To: joseph at codesourcery dot com
- Cc: libc-alpha at sourceware dot org
- Date: Wed, 05 Dec 2012 13:33:13 -0500 (EST)
- Subject: Re: Increase width of padded output in stdio-common/tst-put-error
- References: <Pine.LNX.4.64.1212051735270.1819@digraph.polyomino.org.uk>
From: "Joseph S. Myers" <joseph@codesourcery.com>
Date: Wed, 5 Dec 2012 17:36:09 +0000
> MIPS testing for 2.17 showed, in addition to failures known from 2.16
> or other architectures, and a couple of new failures where the test or
> the relevant bit of it was new but the cause was the same as other
> previously known failures, unexpected failures of
> stdio-common/tst-put-error ("padded fprintf succeeded").
>
> The test checks for errors being detected when writing 10000 bytes of
> padded string with fprintf. This will only occur if the bytes do not
> fit in stdio buffers so an attempt is actually made to write them out.
> The file in question is on /tmp (before the file descriptor is closed
> to cause the errors to occur) and the buffer size is determined by
> libio from the block size for that file. When /tmp is a tmpfs
> filesystem, as on the test systems I'm using for MIPS, the block size
> is the kernel's page size - which is 16384 bytes in this case.
>
> It appears the Linux kernel has various cases with page sizes up to at
> least 1MB. This patch changes the test to use a width of 10000000
> instead of 10000, hopefully covering all the likely cases of larger
> page sizes causing this test to fail.
>
> Tested x86_64 and x86, and on MIPS that this does fix the observed
> failure.
>
> 2012-12-05 Joseph Myers <joseph@codesourcery.com>
>
> * stdio-common/tst-put-error.c (do_test): Use 10000000 instead of
> 10000 as width of padded output.
This is fine.