[diffoscope] diffoscope - Excessive memory usage?

Greg Chabala greg.chabala at gmail.com
Fri Sep 12 01:46:27 UTC 2025


Oh, one more thing, I checked the exit code after running it a second time
and getting no output: 137. Likely killed by docker for exceeding some
default memory constraints, or perhaps it completely exhausted the swapfile
eventually.

On Thu, Sep 11, 2025 at 8:42 PM Greg Chabala <greg.chabala at gmail.com> wrote:

> Hello Chris,
>
> I wonder: is the Docker container using a tmpfs?
>
>
> I don't know if this is a question directed to me, but I ran the command
> as shown on https://diffoscope.org/ , just with the ISOs on my desktop,
> e.g.:
>
> docker run --rm -t -w $(pwd) -v $(pwd):$(pwd):ro
>> registry.salsa.debian.org/reproducible-builds/diffoscope image1.iso
>> image2.iso
>
>
> So if I should start it differently to use (or avoid using) a tmpfs, I
> wasn't really thinking about it.
>
> Cheers,
> Greg
>
> On Thu, Sep 11, 2025 at 8:24 PM Chris Lamb <chris at reproducible-builds.org>
> wrote:
>
>> Hello Greg,
>>
>> > Potentially user error here, but hoping for some ideas.
>>
>> Well, we don't like to assume user error — or putting it another way,
>> diffoscope should make it easy to do the right or expected thing. :)
>>
>> > I pulled the latest diffoscope container image to compare two 4.7GB DVD
>> > ISOs of the same disc. After starting it, it quickly used all 16GB of
>> > RAM on the machine, and then another ~8GB of swap.
>>
>> ISO images are particularly taxing for diffoscope, but it shouldn't be
>> hitting memory quite like that. I wonder: is the Docker container
>> using a tmpfs? That would certainly exhaust your memory very quickly,
>> as the ISO image needs to be extracted before its contents can be
>> compared.
>>
>>
>> Regards,
>>
>> --
>>       o
>>     ⬋   ⬊      Chris Lamb
>>    o     o     reproducible-builds.org 💠
>>     ⬊   ⬋
>>       o
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.reproducible-builds.org/pipermail/diffoscope/attachments/20250911/68adcf29/attachment.htm>


More information about the diffoscope mailing list