Re: Latest Glibc from CVS has segmentation problems.

["Kevin P. Fleming" <kpfleming at backtobasicsmgmt dot com>]

Alfred M. Szmidt wrote:

>    It seems that such reports of errors and omissions in INSTALL are
>    no longer welcome. In fact, I almost wonder if the user community
>    would be better served by replacing the entire contents of INSTALL
>    with "Don't", possibly followed by "And if you absolutely must,
>    copy what the distribution makers do."
>
> They are quite welcome, when someone sends a patch for them, not when
> someone bitches about it without doing any work.  Nobody has even
> tried to update the INSTALL here, or actually suggested improvements
> for it.

This list can be a black hole for patches. I have twice posted a 
correctly formatted patch, against CVS glibc, with a properly formatted 
ChangeLog entry, to fix a very obvious set of errors in a glibc 
testcase. This patch did not introduce new functionality, and did not 
change any functionality that was not obviously broken, because this 
testcase has a dozen or more sub-tests, and the older ones work the 
correct way while the newer ones do not. This was most likely caused by 
a simple cut-and-paste error that was never corrected.

The first time I sent the patch, on 2004-01-27, no response was 
forthcoming. I sent it again on 2004-02-09, and this time Roland 
accepted it and committed, although silently :-) So in spite of all this 
discussion, properly constructed patches that fix actual problems _do_ 
get accepted. There is a hurdle to get over in that you have to make 
sure you conform to the not-at-all-documented requirements for patch 
submission, but lurking on this list for a while will allow you to learn 
what those requirements are. I learned because my first attempts at 
solving the problem were in entirely the wrong direction, and once I 
found the correct direction another problem surfaced. This is the 
problem whose patch was actually accepted (after Uli correctly rejected 
the first one).

Possibly a helpful solution to this problem would be for Uli and the 
others to appoint one or more "lieutenants" (as Linus Torvalds has done) 
to deal with the people who are working below the maintainers' level.