`timsconvert` mzML files unreadable in DIA-NN

[sjust-seerbio]

We have observed an error with DIA-NN 1.8.1 reading certain mzML files created with timsconvert:

E:\EXP24054_2024ms0528X1_A_GA1_1_27896.mzML(1) : parseOffset() 2: Syntax error parsing XML.

Based on the results of running DIA-NN with strace it appears this crash occurs attempting to read the <indexList> element, which is not found at the reported indexListOffset.

The issue with the mzML can be verified using tail:

$ tail -c 200 EXP24054_2024ms0528X1_A_GA1_1_27896.mzML
 idRef="scan=51335">846541370</offset>
    </index>
  </indexList>
  <indexListOffset>846547400</indexListOffset>
  <fileChecksum>2a757d5a71d7aca0788a302e8b881e42d4045446</fileChecksum>

$ tail -c +846547400 EXP21063_2022bruker038bX25_A_BA4_1_757.mzML  | head
ML>
  <indexList count="1">
    <index name="spectrum">
      <offset idRef="scan=1">3832</offset>
      <offset idRef="scan=2">143201</offset>
      <offset idRef="scan=3">149715</offset>
      <offset idRef="scan=4">156293</offset>
      <offset idRef="scan=5">161496</offset>
      <offset idRef="scan=6">167723</offset>
      <offset idRef="scan=7">174251</offset>

The root cause of this error appears to be the update_spectra_count() function, which is not guaranteed to preserve byte offsets in the file.

We will provide a PR patching this issue.