Re: [Mldonkey-devel] RFC : logging in mldonkey

[Fortin Denis]

After introducing the first logging bug, i am beginning to think that it
might no be a good idea to use camlp4 everywhere.
The bug was in BTClients.check_finished : 

===== Buggy =====

#source :

if (file_size file <> Int64Swarmer.downloaded swarmer) then
   BLOG WARN : 
    "%s : Downloaded size differs after complete verification\n"
    (file.file_name) 
   ELOG;
download_finished file

#preprocessed as :
      if file_size file <> Int64Swarmer.downloaded swarmer then
   begin 
     Printf2.lwarn "WARN  : %s : Downloaded size differs after complete
verification\n" file.file_name;
    download_finished file 
   end

-----------------

===== Fixed =====
#source :

BLOG WARN (file_size file <> Int64Swarmer.downloaded swarmer) : 
          "%s : Downloaded size differs after complete verification\n"
          (file.file_name) 
ELOG;
download_finished file

#preprocessed as :
  
if file_size file <> Int64Swarmer.downloaded swarmer then
 Printf2.lwarn "WARN  : %s : Downloaded size differs after complete
verification\n" file.file_name;
download_finished file

-----------------

As you can see in the buggy piece of code the download_finised was
called only when file size differs.
I think that the first one is valid in ocaml and was supposed to do
something right.
But for a reason that i don't understand camlp4 as decided to put a
begin end at the wrong place (download_finished shouldn't be in the
begin end).

I don't know what is wrong in this case : my macro or camlp4 (i guess it
is my macro).
This left me with some questions : 
Should we keep the macro like it is and trust camlp4/my macro in doing a
good job in all files of mldonkey?
Should we dump macros and use Printf2.ldebug (and preprocess only
printf2.ml) and hope that ocaml compiler will remove tests like this
if var=false then (); ?
I don't know what to think about this.


Denis