2026/04/23
Jump to navigation
Jump to search
|
Thursday, April 23, 2026 (#113)
|
|
References |
transcribed retroactively from a Signal chat
- 09:20 ...and now I have to figure out how to set the Updater from a place that isn't supposed to have to do that (now), but pretty sure it's really easy if I can just remember the details (!summon codevætter).
- 11:20 Readout is now adequate, though there are behaviors I still don't quite understand. Running a test now to see if it detects EoF properly.
Apparently it did! Next test is to try importing to remote server over ssh....except it didn't actually get all of the tables. I suspect a logic bug.
- 11:26 Maybe needs a post-loop buffer-flush...
- ...or else defer the loop-exit until the output's bytes-written matches the input's bytes-read. I like that idea better.
- 11:27 Maybe somewhere in this investigation, I'll figure out why it seems to be loading so much into RAM between writes.
- 14:45 ...data not being fully imported. Source file 537.1 MB, re-exported data 315.6 MB. (I supposed it's remotely possible that the export is somehow encoded more efficiently than the original, but that seems unlikely. I guess to be sure, I should to a MariaDB-to-MariaDB comparison (source file is from MySQL, and rather an old version at that).
- 16:30 Came up with a better EoT-detection method and am now... closer, but still encountering a lack of cigar-manifestation. 502.3 MB out of 537.1.
- (End-of-Transmission is probably a better way to think of this: "when is the data done being transmitted", not "when have we read all the data".)
- 16:41 kdiff3 refuses to show the actual contents of the files because "Line end style: Unknown" (which Kate does not seem to have a problem with)
- 16:49 Maybe I just need, like, a command which will list all the tables and how many rows each one has.
- It would make sense to tag that on to DBA's
-l/listcommand.
- It would make sense to tag that on to DBA's
- 16:51 Maybe I should poke at import-over-ssh2, to see how broken that is.
- 16:53 No overt error messages; gets set up nicely, goes into loop -- nothing gets sent, forever. Okay.
- 20:56 So, now, looking at the ssh2 import: it does create the schema, so that's something. Then it apparently manages to send 80 bytes, and then it gets stuck.
- 20:59 Trying to figure out a way to see what those 80 bytes are -- might be a clue.
- 21:29 I wonder if it's having trouble switching directions. First it has to send an in-memory SQL string to the process and wait for some data, then it has to send SQL from a file to the process and keep looping until all of the file is sent.
- Also, obviously there has to be some difference between the local-proc and ssh2 versions, so it's going to be code in those classes... probably/maybe...
- 21:33 I wasn't resetting the streams after setup in ssh2 -- maybe that will make a difference?
- 21:34 No apparent change. Moving on...
- 22:03 Made a progress: the Stream for the source-data was not being set up properly in ssh2. Now getting actual errors! Time to stop, tho.