Re: [rt-tests][PATCH] rt-tests: make manpages builds reproducible

["Jeremy A. Puhlman" <jpuhlman@mvista.com>]

On 2/26/2020 9:18 AM, John Kacur wrote:
>
> On Tue, 25 Feb 2020, Jeremy A. Puhlman wrote:
>
>> From: Jeremy Puhlman <jpuhlman@mvista.com>
>>
>> Add -n to gzip call to make the build output
>> of the manpages reproducible.
>>
>> Signed-off-by: Jeremy A. Puhlman <jpuhlman@mvista.com>
>> ---
>>   Makefile | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/Makefile b/Makefile
>> index 8747971..1b37ba7 100644
>> --- a/Makefile
>> +++ b/Makefile
>> @@ -181,7 +181,7 @@ ssdd: $(OBJDIR)/ssdd.o $(OBJDIR)/librttest.a
>>   	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $< $(LIBS) $(RTTESTLIB)
>>   
>>   %.8.gz: %.8
>> -	gzip -c $< > $@
>> +	gzip -nc $< > $@
>>   
>>   %.8.bz2: %.8
>>   	bzip2 -c $< > $@
>> -- 
>> 2.20.1
>>
>>
> Could you explain to me how this makes the build output of manpages more
> reproducible?

gzip adds the name of the file(not really an issue) and the modification 
time into the header of the
gzipped archive.  Different modification times can cause the archives to 
have different md5sums even
though the content is identical.  Adding -n causes the archives to 
always be identical regardless of when
they are built so long as the content is identical. Its less of an 
issue(provided the build system in question
doesn't touch or modify the man pages) when you are building from 
released tarballs, but if you build
from git, the modification time of the file is when the file was checked 
out.

Tools like rpm will object to the files being installed(like say 
multilib versions of the same packages) together
due different md5sums and the fact that they are not elf binaries, even 
though the files are the same.


-- 
Jeremy A. Puhlman
jpuhlman@mvista.com