Feature request: warn on overwrite

[Andi47]

I just copied my relation to a msieve.dat file, (i'm glad that I *copied* and not *moved* them, made up my worktodo.ini and msieve.fb files and then started msieve (1.39).
When it said "sieving in progress", I recognized that i had forgotten the -nc command - and it had overwritten my msieve.dat file.

What I'd wish to have (just for such cases) is that msieve checks if the files mentioned before exist, that it asks "msieve.dat already exists - overwrite y/n?" (the dos-comand "copy *.out msieve.dat" takes a few minutes when I have got nearly 2 GB of relations)

Edit: Addendum: It would also be nice if msieve checked existing files if they fit to the actual job, and if yes, that it'd append instead of overwriting the msieve.dat file.

[jasonp]

This comes up over and over again.

The problem is that people want msieve to handle batches of numbers, not just one at a time. So how are you supposed to know when an existing savefile is supposed to be appended to, or overwritten completely when you start a new number? This is why the number being factored must be in the first line of the savefile. NFS postprocessing will never overwrite the savefile, but switching from NFS to QS means the savefile is now stale. Asking for user input isn't going to work in scripts. Maybe the binary should just never allow batch input of numbers, that would make restarting easy.

[Andi47]

Originally Posted by jasonp

So how are you supposed to know when an existing savefile is supposed to be appended to, or overwritten completely when you start a new number?

add an -overwrite option which handles batch input without asking for user's input.

[10metreh]

I just suffered a similar fate. I had terminated a factMsieve.pl run (spawned by aliqueit) during the linear algebra, and I restarted aliqueit as usual. First of all aliqueit ran "msieve -np -v" to find a poly - this went straight through as there was already a poly. Then it started factMsieve.pl which started doing the matrix again - except that I found my test.dat (aliqueit equivalent of msieve.dat) file had been overwritten. I have two questions:

1. What caused the overwrite?
2. Can you complete a job when you only have the cycles, not the relations?



And no, I didn't run a QS, and even if I did, it would have used a separate msieve.dat.

[jasonp]

Could aliqueit or the perl script delete existing files before commencing the postprocessing? NFS postprocessing should not delete the savefile for any reason now.

The cycles are not enough to complete the job, you need the relations that generated them.

[10metreh]

Originally Posted by jasonp

Could aliqueit or the perl script delete existing files before commencing the postprocessing? NFS postprocessing should not delete the savefile for any reason now.

The cycles are not enough to complete the job, you need the relations that generated them.

Right then, I'll go and post this over in the aliqueit thread... (Of course it's possible there's a bug in factMsieve.pl, since not many people ever stop the script during a linalg run.)

[10metreh]

It's completed normally:

Sat Jun 06 11:29:51 2009  prp30 factor: 372267511290260300312765522717
Sat Jun 06 11:29:52 2009  prp70 factor: 1546433225670278770208034393844290214075157728093113941172801005679059

I am fed up with P30s turning up in my high C9x GNFSs, though!