This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [patch] Error on setenv(..., NULL, ...)

From: Jeff Law <law at redhat dot com>
To: Roland McGrath <roland at hack dot frob dot com>, Paul Eggert <eggert at cs dot ucla dot edu>
Cc: Paul Pluzhnikov <ppluzhnikov at google dot com>, Szabolcs Nagy <szabolcs dot nagy at arm dot com>, Joseph Myers <joseph at codesourcery dot com>, GLIBC Devel <libc-alpha at sourceware dot org>, "mtk at man7 dot org" <mtk at man7 dot org>
Date: Fri, 13 Mar 2015 11:35:42 -0600
Subject: Re: [patch] Error on setenv(..., NULL, ...)
Authentication-results: sourceware.org; auth=none
References: <CALoOobNSbWUkd_i-L6U0ovbqPYnJY-h=ftX1K61yb19pmJj6aw at mail dot gmail dot com> <alpine dot DEB dot 2 dot 10 dot 1503111712240 dot 30954 at digraph dot polyomino dot org dot uk> <CALoOobPKxfJfnbcUKH8osgCZMeSiD83K1OiF+_vSeAy0ewe1Jw at mail dot gmail dot com> <55008721 dot 1090200 at arm dot com> <CALoOobNbOgm5=oFbEUmTbca3M-KqSUgGmTeWYOt1zTN-CTLoog at mail dot gmail dot com> <55008DE0 dot 8050805 at cs dot ucla dot edu> <20150312215314 dot 1B7CC2C3B8E at topped-with-meat dot com> <55021AB7 dot 1060905 at cs dot ucla dot edu> <20150313170436 dot BF4C92C3B3B at topped-with-meat dot com>

On 03/13/15 11:04, Roland McGrath wrote:

That is fine by me.  I'm a bit surprised the compiler actually figures out
that VALLEN cannot be used uninitialized and doesn't emit the warning.
But great if so.

I haven't looked deeply at this example, but we're consistently lookingto improve GCC's ability to prove certain edges in the control flowgraph are not executable.

Of particular interest are equivalences that are only valid onparticular paths. I can speculate that later in this code there's a useof VALLEN that is conditional on the value of COMBINED.

Through block copying and simplifications we can isolate the pathsleaving the first combined != NULL test. On one of those paths there'llbe the second combined != NULL test which becomes redundant (ie knownstatic result). We eliminate the redundant test, simplify the CFG andas a result the uninitialized use is removed and ultimately better code.


Jeff

References:
- [patch] Error on setenv(..., NULL, ...)
  - From: Paul Pluzhnikov
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Joseph Myers
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Paul Pluzhnikov
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Szabolcs Nagy
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Paul Pluzhnikov
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Paul Eggert
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Roland McGrath
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Paul Eggert
- Re: [patch] Error on setenv(..., NULL, ...)
  - From: Roland McGrath

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]