rnx2rtkp (RTKLIB_EX 2.5.0): malloc error in addobsdata for high-rate RINEX 3 base files

[stefanomoro]

When running rnx2rtkp on a short rover observation window (≈24 minutes) at 10 Hz, using a full-day RINEX 3 base file, the process fails with:

addobsdata: malloc error n=288x8388608
error : insufficient memory

This corresponds to a requested allocation of ~2.3 GB.

The same dataset:

• runs successfully in RTKPOST (GUI), reaching ~2.3 GB of memory
• runs successfully in an older RTKLIB release, demo5L
• runs successfully with -ti 1 (1 Hz)
• fails with -ti 0.1 (10 Hz), even when -ts/-te are set to a short window

Command used :

rnx2rtkp rover.obs base.25O base.25P \
  -ts 2025/12/10 09:35:10 \
  -te 2025/12/10 09:59:10 \
  -k rtklib_ppk.conf \
  -x 3 \
  -o test.pos

PC system spec:

OS: Windows 10 / Windows Server (also reproduced on laptop)
Architecture: 64-bit
RTKLIB build: RTKLIB_EX 2.5.0
Previous version tested: demo5L
GUI vs CLI: RTKPOST (GUI) works, rnx2rtkp (CLI) fails

Expected behavior:

• ts/-te should prevent rnx2rtkp from allocating buffers proportional to the full base file when a short time window is requested, or
• memory usage should be comparable to RTKPOST or previous releases for the same dataset.

Please let me know if this is expected behavior, a known limitation, or a regression between releases. I’m happy to test patches or provide additional diagnostics.

[ourairquality]

The start and end time should be used here, see rinex.c readrnxobs() and the call to screent(). That reducing the interval helps suggests that this is working for at least the interval. Not clear why the start and end time are not working here. Can you debug that to see why it is failing? Did a quick check here on some data, 1Hz, and the code was working, the number of epochs used was reduced to the time span.