This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC 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/Merge Request] Vtable Verification feature.


On Tue, Aug 6, 2013 at 12:07 PM, Caroline Tice <cmtice@google.com> wrote:
>
>
>
> On Tue, Aug 6, 2013 at 12:02 PM, Ian Lance Taylor <iant@google.com> wrote:
>>
>> On Tue, Aug 6, 2013 at 11:43 AM, Caroline Tice <cmtice@google.com> wrote:
>> >
>> > On Tue, Aug 6, 2013 at 11:31 AM, Ian Lance Taylor <iant@google.com>
>> > wrote:
>> >
>> >>The output to the file doesn't have
>> >> any indication of what file is being compiled, so it will be ambiguous
>> >> when run in parallel.
>> >
>> > You are mistaken.  It outputs one line to the log file for each
>> > compilation
>> > unit.  The output line begins with the name of the file that was being
>> > compiled.  In my use case, I have used this to build a very large
>> > system,
>> > which resulted in something like at 8000 line log file of counts, which
>> > I
>> > then used my sum script on to see how the verifications were going.
>>
>> I was mistaken in detail but I'm not sure I was mistaken in principle.
>> What happens if you are building the large system twice in different
>> directories on the same machine?  Or, for that matter, if two
>> different people are doing so?  Or if one person did it a while ago,
>> and now you want to do it, but you can't open the file because it is
>> owned by the other person?
>>
>> Maybe you should simply change -fvtv-counts to take a file name, then
>> we don't have to worry about any of this.
>>
> It's not quite that simple:  the -fvtv-counts flag actually causes two files
> to be created; also there's another flag, -fvtv-debug that generates a third
> file (i wanted a lot of information when I was working on and debugging this
> feature).  Also if users are arbitrarily allowed to name the counts file
> anything, the summing script program I wrote won't be able to find them.

That doesn't seem like a compelling argument to me, since one could
pass the file names to the summing script as well.

As far as I can see, on a multi-user system, there is no reasonable
alternative to permitting the user to specify the file names to use,
or at least a directory where the files should be placed.  And if
permit that, why not simply require it?

Ian


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