Re: [Dragora-users] tde-i18n long package processing

[DustDFG]

On Sat, Jul 16, 2022 at 3:48 PM Antonio Diaz Diaz <antonio@gnu.org> wrote:
>
> DustDFG wrote:
> >>> The unpacking of tde-i18n archive takes long time (in my case ~10
> >>> minutes). As I know, you prefer to use for tarlz "--solid -9" options
> >>> and I think that you used it with this package. If it is so, I want to
> >>> propose to use for this package other compression options that can
> >>> speed up unpacking without significant damage for archive size
> >
> > As I know, you can find its sources[1][2] at dragora rsync server [3]
> > and mirrors of the rsync server.
> >
> > [1] 
> > http://mirror.cedia.org.ec/dragora/current/sources/tde-i18n-20220504_e67bfc8629.tar.lz
>
> Thank you.
>
> I have downloaded the file and it seems that Matías is already using optimal
> options to create it: '-9 --bsolid'. You can verify this with the following
> command:
>
> $ lzip -lvv tde-i18n-20220504_e67bfc8629.tar.lz
>     dict   memb  trail   uncompressed     compressed   saved  name
>    32 MiB    17      0     1035493376      210839116  79.64%
> tde-i18n-20220504_e67bfc8629.tar.lz
>   member      data_pos      data_size     member_pos    member_size
>       1              0       67124736              0        8733513
>       2       67124736       67121664        8733513       17127777
>       3      134246400       67273728       25861290       12500459
>       4      201520128       67140608       38361749        7175227
>       5      268660736       67115008       45536976       10444313
>       6      335775744       67115520       55981289       28148400
>       7      402891264       67111424       84129689       21878554
>       8      470002688       66850304      106008243        9424019
>       9      536852992       67107840      115432262       12988451
>      10      603960832       67101696      128420713       11146566
>      11      671062528       67110912      139567279       24369671
>      12      738173440       67100160      163936950        6647518
>      13      805273600       67116032      170584468        6911856
>      14      872389632       67121152      177496324       15048311
>      15      939510784       67098624      192544635       10330287
>      16     1006609408       28882944      202874922        7964150
>      17     1035492352           1024      210839072             44
>
> On a dual core machine it takes tarlz 11.2 seconds to list the archive,
> against 19.7 seconds for GNU tar. Almost twice as fast. The same happens
> when extracting to a tmpfs on RAM, 12.1s for tarlz vs 21.6s for GNU tar.
>
> I'm afraid the reason why it takes so long to extract on your machine is
> that it is slow to write so many files to your storage (56745 files and
> directories totalling more than 1 GB). Unless there is some bottleneck in
> tarlz that I don't know about. Please, could you try to extract the archive
> with tarlz and GNU tar and compare the times? Thanks.
>
> Best regards,
> Antonio.
>
>

Hello Antonio!

I didn't try to know exact time of extraction but I got strange
results when I tried to get exact time of listing the archive.

The followinfg results for standart tar:

$ time tar -t -f ./tde-i18n-20220504_e67bfc8629.tar.lz

real    0m27.391s
user    0m26.549s
sys    0m1.102s

The followinfg results for tarlz without -f option:

$ time tarlz -t ./tde-i18n-20220504_e67bfc8629.tar.lz

real    56m5.702s
user    0m0.001s
sys    0m0.006s

I couldn't stand my patience so I interrupted process with CTRL+C

But the prevoius command with "-f" flag:

$ time tarlz -t ./tde-i18n-20220504_e67bfc8629.tar.lz

real    0m11.626s
user    0m34.261s
sys    0m0.510s