(32-bit compilation) ipc/chromium/src/third_party/libevent/buffer.c:2518:18: warning: comparison is always false due to limited range of data type [-Wtype-limits]

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

(32-bit compilation) ipc/chromium/src/third_party/libevent/buffer.c:2518:18: warning: comparison is always false due to limited range of data type [-Wtype-limits]

ISHIKAWA,chiaki
Hi,

Is building mozilla code such as firefox and thunderbird under 32-bit linux
a supported operation?

During compilation of C-C TB under 32-bit linux,
I think I found a serious issue regarding the mismatch of the size
of literal constant and the size of variables (or structured member)
which renders some run-time check ineffective and thus run-time
error conditions are not handled correctly.

Usually, I have to ignore the signed/unsigned comparison, etc. during build
since there are just too many, but this time
the particular warning about the size mismatch against a member
field/variable (|st_size|) of a structure passed to |fstat| caught my eye,
and I noticed the possible seriousness of the issue if my understanding is
correct.


The system is Debian GNU/linux 32-bit:
uname -a
Linux debian-vbox-ci 3.2.0-4-686-pae #1 SMP Debian 3.2.68-1+deb7u1 i686
GNU/Linux

I understand that for official distribution,
mozilla compilation farm seems to use cross compiling from the 64-bit
linux to produce 32-bit linux version of firefox and thunderbird.
If so, then maybe for the official version, the build may happen to pick up
the right size accidentally. Frankly I don't know what is going on,
I suspect there is something incorrect in the headers regarding the behavior
that ought to change according the target/host environment.

Eveb if the official version happens to be correct (by mistake?), still I
want the headers to be correctly written to avoid future issues.

Also, the problem occurs in mozilla/ipc/chromium/src/third_party/libevent
directory.
Yes, I noticed "/third_party/" moniker, but the compilation warning
and the resulting failure of correct runtime check
looks serious enough to discuss the remedy in mozilla community in the
first place. If a consensus patch emerges, we can send the patch
upstream.

TIA
_______________________________________________
dev-builds mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-builds
Reply | Threaded
Open this post in threaded view
|

Re: (32-bit compilation) ipc/chromium/src/third_party/libevent/buffer.c:2518:18: warning: comparison is always false due to limited range of data type [-Wtype-limits]

Mike Hommey
On Mon, Jul 06, 2015 at 06:44:45PM +0900, ishikawa wrote:

> Hi,
>
> Is building mozilla code such as firefox and thunderbird under 32-bit linux
> a supported operation?
>
> During compilation of C-C TB under 32-bit linux,
> I think I found a serious issue regarding the mismatch of the size
> of literal constant and the size of variables (or structured member)
> which renders some run-time check ineffective and thus run-time
> error conditions are not handled correctly.
>
> Usually, I have to ignore the signed/unsigned comparison, etc. during build
> since there are just too many, but this time
> the particular warning about the size mismatch against a member
> field/variable (|st_size|) of a structure passed to |fstat| caught my eye,
> and I noticed the possible seriousness of the issue if my understanding is
> correct.
>
>
> The system is Debian GNU/linux 32-bit:
> uname -a
> Linux debian-vbox-ci 3.2.0-4-686-pae #1 SMP Debian 3.2.68-1+deb7u1 i686
> GNU/Linux
>
> I understand that for official distribution,
> mozilla compilation farm seems to use cross compiling from the 64-bit
> linux to produce 32-bit linux version of firefox and thunderbird.
> If so, then maybe for the official version, the build may happen to pick up
> the right size accidentally. Frankly I don't know what is going on,
> I suspect there is something incorrect in the headers regarding the behavior
> that ought to change according the target/host environment.
>
> Eveb if the official version happens to be correct (by mistake?), still I
> want the headers to be correctly written to avoid future issues.
>
> Also, the problem occurs in mozilla/ipc/chromium/src/third_party/libevent
> directory.
> Yes, I noticed "/third_party/" moniker, but the compilation warning
> and the resulting failure of correct runtime check
> looks serious enough to discuss the remedy in mozilla community in the
> first place. If a consensus patch emerges, we can send the patch
> upstream.

It's a mozilla only problem in event-config.h for linux and mac. Please file a bug.

Mike
_______________________________________________
dev-builds mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-builds
Reply | Threaded
Open this post in threaded view
|

Re: (32-bit compilation) ipc/chromium/src/third_party/libevent/buffer.c:2518:18: warning: comparison is always false due to limited range of data type [-Wtype-limits]

ISHIKAWA,chiaki
On 2015/07/07 6:32, Mike Hommey wrote:

> On Mon, Jul 06, 2015 at 06:44:45PM +0900, ishikawa wrote:
>> Hi,
>>
>> Is building mozilla code such as firefox and thunderbird under 32-bit linux
>> a supported operation?
>>
>> During compilation of C-C TB under 32-bit linux,
>> I think I found a serious issue regarding the mismatch of the size
>> of literal constant and the size of variables (or structured member)
>> which renders some run-time check ineffective and thus run-time
>> error conditions are not handled correctly.
>>
>> Usually, I have to ignore the signed/unsigned comparison, etc. during build
>> since there are just too many, but this time
>> the particular warning about the size mismatch against a member
>> field/variable (|st_size|) of a structure passed to |fstat| caught my eye,
>> and I noticed the possible seriousness of the issue if my understanding is
>> correct.
>>
>>
>> The system is Debian GNU/linux 32-bit:
>> uname -a
>> Linux debian-vbox-ci 3.2.0-4-686-pae #1 SMP Debian 3.2.68-1+deb7u1 i686
>> GNU/Linux
>>
>> I understand that for official distribution,
>> mozilla compilation farm seems to use cross compiling from the 64-bit
>> linux to produce 32-bit linux version of firefox and thunderbird.
>> If so, then maybe for the official version, the build may happen to pick up
>> the right size accidentally. Frankly I don't know what is going on,
>> I suspect there is something incorrect in the headers regarding the behavior
>> that ought to change according the target/host environment.
>>
>> Eveb if the official version happens to be correct (by mistake?), still I
>> want the headers to be correctly written to avoid future issues.
>>
>> Also, the problem occurs in mozilla/ipc/chromium/src/third_party/libevent
>> directory.
>> Yes, I noticed "/third_party/" moniker, but the compilation warning
>> and the resulting failure of correct runtime check
>> looks serious enough to discuss the remedy in mozilla community in the
>> first place. If a consensus patch emerges, we can send the patch
>> upstream.
>
> It's a mozilla only problem in event-config.h for linux and mac. Please file a bug.
>
> Mike

Thank you for the comment.

Will do.

CI




_______________________________________________
dev-builds mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-builds