Re: 10 times performance degradation for search in large tar in 1.26+

Hello Eugene,

Thank you for your report.

Feel free to create a PR on GitHub (with tests) so we can see what this
would look like.

Thank you,
Gary

On Thu, Aug 6, 2026, 09:53 Okhotnikov, Evgenii via user <
[email protected]> wrote:

> Hi
>
> I use common-compression to with with large tar files (~50 GB). I saw huge
> performance degradation (10 times) when search for particular file using
> recent version (starting from 1.26) compared to 1.25 and below . the issue
> is compression IOUtils.skip now uses common-io IOUtils skip. old
> implementation was using skip on underlying stream to skip as much as
> possible, then read the remaining data. now, common-io IOUtils reads all
> the data instead of skipping it.
>
> It makes searching in large TAR about 10 times slower and saturates my SSD
> IO. I use buffered stream, and huge buffer size helps a bit - it makes it 5
> times slower instead of 10 times. I'd like to go back to old behavior and
> just skip the entries I don't need instead of reading them , but there's no
> proper way to modify the behavior in my app itself - skip is called from
> private methods
>
> My proposal is to make it configurable in TarArchiveInputStream  ,
> common-compression IOUtils or even common-io Utils how to skip - with pure
> read or with skip and read combination, or add protected method
> skip(InputStream in, long bytesToSkip) method in TarArchiveInputStream so
> that I could extend the class and modify the behavior.
>
> Eugene
>
> CONFIDENTIALIT? : Ce document est destin? uniquement ? la personne ou ?
> l'entit? ? qui il est adress?. L'information apparaissant dans ce document
> est de nature l?galement privil?gi?e et confidentielle. Si vous n'?tes pas
> le destinataire vis? ou la personne charg?e de le remettre ? son
> destinataire, vous ?tes, par la pr?sente, avis? que toute lecture, usage,
> copie ou communication du contenu de ce document est strictement interdit.
> De plus, vous ?tes pri? de communiquer avec l'exp?diteur sans d?lai et de
> d?truire ce document imm?diatement.
> CONFIDENTIALITY: This document is intended solely for the individual or
> entity to whom it is addressed. The information contained in this document
> is legally privileged and confidential. If you are not the intended
> recipient or the person responsible for delivering it to the intended
> recipient, you are hereby advised that you are strictly prohibited from
> reading, using, copying or disseminating the contents of this document.
> Please inform the sender immediately and delete this document immediately.
>
>