Memory leak?

[Selur]

Is it just me? After running media-autobuild_suite ~50% (32GB) of my memory, seem to be locked by something and the only way to get them back is to reboot my system.
Before trying to figure out more about this, I was wondering if I'm the only one experiencing this. :)

[L4cache]

Uh, no?
Is this happens to you every time?

[1480c1]

I haven't experienced this, so I don't have much of a clue here. Did task manager not show any program with outstanding memory usage?

[hydra3333]

I run it in a windows 11 sandbox, with a link defined to a host's folder to save stuff onto.
Works a treat, everything cleaned up after the sandbox closes, no dangling registry entries or temp folder files, etc etc. Safe as houses :) Let me know if you want me to post my example .wsb and .bat to set it up before running MABS.
Cheers

[Selur]

Got no problem with files, just after running MBAS nowadays, half my ram is occupied.
I'll do some testing in the next few weeks to see whether this is just caused by a specific build. :)
Happy this does not happen for others, will report back when I know more. :)

---

No neither ressource monitor nor taskmgr show where the RAM is.

[Selur]

From the looks of it, each compile step increases the used memory (both during 32bit and 64bit) and does not free it.
When running the script and then restarting because of updates and thus rerunning, I end up with 48% memory usage and not trace where they went. Will do more testing and report back again.

[L4cache]

You can try search for the "root path" of your m-ab-s (e.g. C:\m-ab-s) in the "Associated Handles", "CPU" tab of the Resource Monitor
This won't tell you where the memory is, but it will show you basically any process related to the suite.

[Selur]

hmm I see, bash.exe, explorer.exe, git, mitty.exe each two times, and then a bunch of sh.exe and make.exe calls.

(none of the sed, mv, gcc processes are listed in the filter)
I'll see whether there is anything left, but I expect not, since I didn't find anything before.

[Selur]

=> handles left, but 35% of my RAM is 'missing', tried using https://learn.microsoft.com/de-de/sysinternals/downloads/rammap and calling "Empty->Working sets" does seem to free a large chunk (~30%) of the memory.

[Selur]

Looking at the Process listing in RamMap there are tons of sed, sh, yasm, bash, etc. processes listed. (all small, but a lot)

[L4cache]

That must be some weird bug, they should exit after build completes...
Maybe a fresh install can fix that, just copy the ini and custom ffmpeg/mpv options and you are good to go.
Just out of curiosity, can you share the whole suite currently on you drive? (there maybe some personal informations inside so be careful though)

[Selur]

Sure, will upload it later and post a link here.

[Selur]

Here you go: https://drive.filejump.com/drive/s/wPDC5EJHLxKJz5Xt19cZvvEoKEyILZ
Will try a fresh install tomorrow.

[Selur]

Let a fresh run overnight and 81% of RAM is used, just let it run again since it reported updates at the end:

Here's a link to the rammap file, created during the second run: https://drive.filejump.com/drive/s/gFMNCCOLkXQGGergkYtFamTxBViWsB
will restart my system once the second pass is over to get some free ram. :)

[L4cache]

I tried to run the copy of your suite (in a VM) and found no memory leak or lingering process, maybe it's some system setting or potentially bugs...

[Selur]

Thanks for testing. At least I know it's something related to my setup and not a general issue. :)

[Selur]

Didn't change a thing today, running the script didn't cause the memory issue. But only a few things were build.
Will report back in case I figure out what causes this here.

[woot000]

A while ago I had a similar issue that affected my entire system -- thousands of zombie processes from the suite were taking up RAM and couldn't be cleared without rebooting Windows.

In my case, I eventually figured out it was my AMD graphics drivers that were the cause -- 24.1.1 through 24.6.1 all regularly left behind zombie processes whenever any program closed. Updating to 24.7.1 and beyond fixed this for me, maybe it's a similar issue for you?