[0/2] MIPS: IRIX ELF format symbol table sort/section symbol name fixes

Message ID alpine.LFD.2.21.2007221254330.234068@eddie.linux-mips.org
Headers show
Series
  • MIPS: IRIX ELF format symbol table sort/section symbol name fixes
Related show

Message

Maciej W. Rozycki July 27, 2020, 9:47 a.m.
Hi,

 Here is a patch series I have prepared as the outcome of a failure with 
MIPS targets using the IRIX rather than traditional ELF format observed 
with a new test case added in the course of addressing PR binutils/25662: 
<https://sourceware.org/pipermail/binutils/2020-March/000336.html>.  The 
test no longer fails for any MIPS targets with these patches in place.

 The changes involve hooks being added to the generic ELF linker and given 
the IRIX ELF format being a dead end, although also being used for several 
non-IRIX MIPS targets, I have decided to keep the changes made to generic 
code to the minimum.  This results in an extra hook, which could possibly 
be otherwise avoided either by involving the relevant existing hook, or by 
redesigning it to serve two related purposes at once, although that would 
be at the cost of rewriting parts of generic code.

 Rainer: You were the reporter of one issue discussed here and partially 
resolved: <https://sourceware.org/ml/binutils/2004-02/msg00072.html> on a 
previous occasion, so I have cc-ed you in case you wanted to know (as a 
matter of interest presumably, given your desire to drop IRIX support from 
binutils).

 Regression-tested across my usual set of 191 targets including 55 MIPS 
ones.  OK to apply?

  Maciej

Comments

Nick Clifton via Binutils July 27, 2020, 12:56 p.m. | #1
On Mon, Jul 27, 2020 at 10:47:52AM +0100, Maciej W. Rozycki wrote:
> Hi,

> 

>  Here is a patch series I have prepared as the outcome of a failure with 

> MIPS targets using the IRIX rather than traditional ELF format observed 

> with a new test case added in the course of addressing PR binutils/25662: 

> <https://sourceware.org/pipermail/binutils/2020-March/000336.html>.  The 

> test no longer fails for any MIPS targets with these patches in place.

> 

>  The changes involve hooks being added to the generic ELF linker and given 

> the IRIX ELF format being a dead end, although also being used for several 

> non-IRIX MIPS targets, I have decided to keep the changes made to generic 

> code to the minimum.  This results in an extra hook, which could possibly 

> be otherwise avoided either by involving the relevant existing hook, or by 

> redesigning it to serve two related purposes at once, although that would 

> be at the cost of rewriting parts of generic code.

> 

>  Rainer: You were the reporter of one issue discussed here and partially 

> resolved: <https://sourceware.org/ml/binutils/2004-02/msg00072.html> on a 

> previous occasion, so I have cc-ed you in case you wanted to know (as a 

> matter of interest presumably, given your desire to drop IRIX support from 

> binutils).

> 

>  Regression-tested across my usual set of 191 targets including 55 MIPS 

> ones.  OK to apply?


The patches look OK to me.

-- 
Alan Modra
Australia Development Lab, IBM
Maciej W. Rozycki July 29, 2020, 8:02 p.m. | #2
On Mon, 27 Jul 2020, Alan Modra wrote:

> >  Regression-tested across my usual set of 191 targets including 55 MIPS 

> > ones.  OK to apply?

> 

> The patches look OK to me.


 Thanks, committed.

  Maciej