bug-apl
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-apl] Fwd: Re: Bogus compiler warnings


From: Christian Robert
Subject: Re: [Bug-apl] Fwd: Re: Bogus compiler warnings
Date: Fri, 10 Mar 2017 02:54:18 -0500
User-agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0

well, no,

I'm perfectly ok with the fact that gnuapl do not compile on old redhat 4, or 
on old redhat 5
with very *old* compilers.

please Upgrade.

my 2 cents,

Xtian.


On 2017-03-10 02:09, Alexey Veretennikov wrote:
Hi Juergen,

Here is the patch which introduces the new configure option:
--disable-werror
If this option is set then the -Werror flag is not added to compiler
options.
By default it is like it is now - -Werror is present.

With this option I was able to turn off compiler warnings.

I ran autoreconf on my machine but it produces more diffs, which I don't
want to attach. I believe it is better to you to run it on your machine
and submit to svn instead of me providing them as a patch.





Juergen Sauermann <address@hidden> writes:

Hi Alexey,

I believe it is not good if we have to introduce additional code just to 
suppress warnings.
In particular if the warning is obviously bogus.

-Werror is on purpose in order to improve the code quality of GNU APL.
If we would turn it off then that goal would be undermined.

I suppose that you can run configure like this:

CXXFLAGS="-Wnoerror" ./configure

If that doesn't work then configure.ac line 39 is place to change the CXXFLAGS.
After that autoreconf and a new ./configure is needed for the change to have an 
effect.

/// Jürgen

On 03/09/2017 11:29 AM, Alexey Veretennikov wrote:

 Hi,

 I'm trying to compile GNU APL with GCC 3.4.6 armv5. There are couple of bogus 
compiler warnings,
 mainly about variable might be uninitialized.
 Sometimes the situations are like this:

 SomeObject * ptr = new SomeObject(arg);

 - here compiler complains (sometimes) what ptr might be used uninitialized.
 I've fixed it by separating variable declaration and initialization with the 
real value, like this:
 In the beginning of the function:

 SomeObject * ptr = 0;

 ....
 ptr = new SomeObject();

 So I can fix them by rearranging a code a little. It doesn't hurt as it seems.
 But maybe it is not way to go, to complicate the code to satisfy the ancient 
compiler?

 There are other issues, mainly related to HEX macro, where we supply pointer 
to this macro.
 On 32bits machine the conversion int64_t(const char*) produces the warning.
 I could solve them by introducing
 HEX(reinterpret_cast<int64_t>(ptr))

 It makes code compile without warnings but looks not that pleasant.

 But maybe there is a possibility to turn off -Werror with the ./configure 
option?

 Br,
 /Alexey






reply via email to

[Prev in Thread] Current Thread [Next in Thread]