What we shipped, in plain language.

Corrections to the Purchase Order print layout so the preview matches the printed document.

The PO History page was rebuilt: the procure-to-pay journey now reads as a single traceable timeline with the per-line matrix and the ordered / received / invoiced / paid reconciliation presented alongside it, and a stage that could not be read says so rather than showing a zero that reads as fact.

Corrections across the cutting room screens — cutting job detail, bundle labels, QC inspection, cutting notes, lay slips and master bags — including the master bag detail rebuild and the data behind the listings.

Open items can be cleared partially, with a clearing amount per invoice; entering none keeps the previous first-in-first-out behaviour. The page previews the exact journal before posting, built by the same code that posts it. The payment advice now quotes what actually reaches the vendor rather than the gross figure, and lists the charge lines. A reversal mirrors the charges.

Fixes to the account and document pickers on the Journal Entry form, and to the Payment Register listing.

A payment can now carry GL lines that are neither the payable nor the bank — bank charges, rounding, an amount withheld — each naming its own account and direction. A charge borne by us clears the invoice in full and pays more out of the bank; a withheld amount clears the invoice in full and pays less. The bank leg is the balancing residual, derived and never typed, so the document cannot post unbalanced whatever is entered.

An export receipt was debited to the bank at amount × the rate somebody typed on the form. The bank converts at its own rate, so that booked the bank for money it never credited and the gap had nowhere to go but the receivable. The rupee figure from the bank advice can now be entered on the receipt: the bank is debited with it, each invoice is still relieved at its own booking rate, and the difference is posted as the realized exchange gain or loss it actually is. Leaving it blank behaves exactly as before.

Where a bank credited amount is entered, the rate it implies — not the header rate — prices every row of the settlement working, and the two legs reconcile to the cent. Where a receipt settles its open items exactly, the rounding residue is absorbed on the last row rather than being stranded as an advance against no invoice.

Deleting a line worked, but the line reappeared on the next load and was still counted in the document total — one requisition served six lines for five live ones and totalled ₹5,02,000 instead of ₹4,72,000. Removed lines are now filtered everywhere they are read.

A service line billed against a posted Service Entry Sheet was reported as “invoiced, goods not received” for the full invoice value. A service is never receivable into stock and an asset receipt is its own document, so neither can produce a goods movement — the report was looking only at goods movements. Each line now resolves its received quantity from the document that actually receives it, so the report ties back to the GR/IR account.

The goods receipt correctly valued a discounted line at its net price, but the vendor invoice proposal offered the gross price, so the invoice billed more than the receipt had accrued and left a permanent unexplained residue in the GR/IR clearing account. The proposal, the price-difference check and the reporting all now read the net price.

The stored requisition total was always missing the most recently added line, which is why every single-line requisition stored zero. One live requisition held ₹4,35,072.75 against seven lines worth ₹5,67,882.00. The total is now recomputed including the line being saved.

Measured on a live workspace with 292 orders and 1,710 lines, the Purchase Order list went from 4,825 database queries and a 1.86 MB response to 2 queries and 176 KB — a total of 10.4 seconds down to 0.9. The list no longer ships every line of every order; an expanded row fetches the one order it opens. Header totals are computed once rather than five queries per row.

The paid data-grid licence key was the last plaintext credential in the source tree, readable in the repository and in build logs. It is now supplied by the build system as a managed secret. Stated plainly rather than left implied: this is a client-side licence, so the value still ships inside the browser bundle — that is how the product is licensed, and the change removes the repository and build-log exposure, not the bundle one.

Observation showed the main application needs no relaxation of the script policy, while a number of older server-rendered pages still carry inline scripts. The strict policy is applied to the application and the relaxed one only to those legacy pages, so a new page inherits the strict policy automatically rather than being missed off a list.

Server-rendered pages pulled executable JavaScript and CSS from four public content delivery networks, with no integrity checking, and at least one reference was unpinned — it resolved to whatever the current release happened to be. A compromised package would have executed in a logged-in user’s session with no deployment on our side. All of these assets are now served from our own static files, removing the third party entirely.

The production planning layout file was four separate templates concatenated into one, so it failed to compile and nothing in that section could be displayed; the schedule dashboard, production order list and production order detail pages additionally had no template at all. All three now render.

Browsers post every kind of report — deprecation, intervention, crash — to the single endpoint they are given. Those were correctly discarded, but recorded as a payload that could not be read, so the status line warned about a problem that did not exist. The line that distinguishes “the policy is clean” from “nothing is arriving” is now trustworthy.

Production orders and their operations are shown on a Gantt chart with their dependencies, and the scheduling dashboard was rebuilt around it.

A new CRP report loads planned and released production against each work centre’s available capacity per period, so an overloaded work centre is visible before the schedule is committed rather than after it slips.

The effect of the factory calendar — holidays and non-working days — on scheduled production is calculated and reported, so a plan is not quietly assumed to run through a shutdown.

The automated web security baseline scan now has a checked-in, reproducible configuration, so a scan result can be compared with the previous one rather than re-derived each time.

Three live credentials were found in the source tree, including a database password written into the build log of every build and the standard administrator password every new workspace was provisioned with — one password, all workspaces, published in the code. All are now supplied from the environment or a managed credential store, and the retired signing key literal was replaced with an obvious placeholder.

The process map read the wrong lifecycle field for Opportunities, which failed the whole case list. The field is now resolved per record type.

Rows on the Pending Goods Receipt grid appeared only after something forced an internal refresh — changing the page size, or visiting the pivot tab and coming back. The grid now refreshes once its rows arrive.

The report-only Content Security Policy shipped with the security hardening work carried nowhere to send reports, so nothing was ever collected and an empty result was indistinguishable from a clean policy. Violations are now received, deduplicated and summarised, in both the formats browsers use, so the observation week the hardening plan depends on can actually run.