AdawatHow we test our tools
What “tested” means here
A tool is not finished when it works on one friendly file. It is finished when it also copes with the awkward ones, behaves the same in Arabic and in English, and says plainly what it cannot do. This page describes the checks we run, so you can judge how far to trust a result and tell us when we got something wrong.
The files we test with
For PDF tools we keep sample documents of 3, 5 and 200 pages, with mixed page sizes and rotations, plus three deliberately bad files: a password-protected PDF, a corrupted one and a text file that only pretends to be a PDF. For image tools we use JPG, PNG and WebP pictures, a phone photo whose rotation is stored in its metadata, a very large 3000 × 2000 px picture and an iPhone HEIC sample. Word conversion is tried with one Arabic and one English document. A tool should either give a correct result or a clear message for each of them.
Automatic checks
Two kinds of tests run before a change goes live. Unit tests check the logic on its own: page order, the search that finds a target file size, colour maths, random draws, date and number formatting. End-to-end tests open the built site in a real browser, add the sample files, press the buttons and inspect the downloaded results, in both languages. After each build, further checks look for broken internal links, pages that cannot be indexed, pages with too little text and pages that repeat each other, and they fail the build if they find any.
Browsers and devices
Every run uses desktop Chromium (the engine of Chrome and Edge) and a phone profile modelled on a Pixel 5, so layouts are checked at phone width as well. We also run the page-navigation and scrolling tests on WebKit, the engine of Safari. We have not tried every phone, tablet or browser version, and a browser we do not mention may behave differently. If something fails for you, tell us which one you use.
Numbers you see on tool pages
When a page quotes a size, such as how small a picture gets when compressed to 100 KB, it comes from running the real tool on a fixed set of generated test images, with the date and method stated next to the numbers. They are not photographs and not promises: your own files will land somewhere else, which is why the pages explain what drives the result.
What we commit to
We do not claim legal validity for anything a tool produces: the signature tool adds a visual signature, not a certified one, and says so. We state a tool’s limits on its page. A tool never silently makes a result worse: the compressor, for instance, returns your original unchanged when it is already under the target. And we fix reproducible bugs, recording the fix in the changelog.
Reporting a problem
Write to contactus@adawats.app with the tool and the steps you took, your browser and device, and the type and rough size of the file (please do not attach private documents; we never need them). A screenshot helps. We usually reply within 2–3 business days, and when we can reproduce the problem the fix is listed in the changelog with its date.